SEM Series (2016) 5. Confirmatory Factor Analysis Part 2
HTML-код
- Опубликовано: 21 авг 2024
- I made a correction to this video and the related stats tools. There was an error in the calculation of the CR. Here is the new video: • SEM Series (2016) 5c. ...
The new video only covers the validity issue though, not all the other topics and procedures covered in this video.
Here's a fun pet project I've been working on: udreamed.com/. It is a dream analytics app. Here is the RUclips channel where we post a new video almost three times per week: ruclips.net/channel/UCiujxblFduQz8V4xHjMzyzQ
Also available on iOS: apps.apple.com/us/app/udreamed/id1054428074
And Android: play.google.com/store/apps/details?id=com.unconsciouscognitioninc.unconsciouscognition&hl=en
Check it out! Thanks!
Dear James, thank you very much for all your videos and information on your website, they are extremely useful! But I'd like to ask a question about the common method variance text with the use of marker variable. According to Podsakoff et al. (2003)'s article, no common latent factor was added to the model with a marker variable. However, the common latent factor is included in your model with marker variable. Could I ask what is the difference please?
@@freebirdavis1 Using the marker extracts the variance common to that marker variable, but not other variance. The common latent factor extracts any other shared variance across all observed items. This is optional and may be argued to be unnecessary. It may even be harmful if we expect all observed items to share trait variance (rather than just method variance).
@@Gaskination Thank you very much for your prompt clarifications! I really appreciate it! And I watched almost all of your videos, you are my life saver! :)
Hi Respected James Gaskin.. Greeting and alot of love from Pakistan. Im gulnaz, phd scholar. I wanna says thanks to you. actually, my field of investigation is management and i have very little knowledge statistical tools. but your videos help me alot to overcome my weak points. I learn SEM only from you.. Islam says if anyone teach you a single word, he/she is your teacher and teachers are like your soul parents. I really respect you alot alot alot. you know, research is Pakistan is on very initial stage and we have little opportunities to learn because of few experts. I really have bundle of love for you. i am big fan of your intellgency level.. You are super cool. keep it up ... you are inspiration of young researchers
I can relate to every word you have written
Please allow me to post these timestamps in this comment. Thank you.
1:53 validity and reliability check
4:55 interpreting validity and reliability results
6:10 common method bias
15:10 imputing factor scores
Dearest Dr. James,
I read a comment that "If the common method bias test indicates that there is no common method bias, then we can go ahead and get rid of the common latent factor and then impute all the factor scores, which will be used in the structural model".
In such an event, can we choose to not impute factor scores and use the full latent model (with latent factors and their respective indicators attached to them) instead?
Infinite thanks to you for making us research scholars' lives easier with your tutorials.
Sincerely, Tushar
Yes, using a latent model is perfectly fine.
@@Gaskination Note taken. I cannot thank you enough for your guidance, Dr. Gaskin! 🙏🏼
Dear James,
I REALLY REALLY appreciate your support. Your Videos and the EXCEL tool is such a great way to get my analysis for my master thesis done. Thank you so much! I will get back to you asa it got marked if you are interested ;)
Clarity to the core ....
Is there something wrong with the Model fit measures plugin?
For example, I got the following results running a model:
CMIN = 228.853
DF = 23.0
CMIN/DF = 9.950 (but the table shows "excellent")
CFI = 0.92 (acceptable) OK
SRMR = 0.093 (table shows "excellent", but it should be "acceptable")
RMSEA = 0.143 (table displays "excellent" but it should be "terrible")
PClose = .000 (need more DF) OK
Why the interpretation of the CMIN/DF, SRMR and the RMSEA is wrong?
I've checked this with other models and apparently there is something wrong with the interpretations done by the plugin.
This is very good to know. Thank you for reporting it. I will pass this along to John (who developed it) and he will check it out. Thanks!
Hi, Prof. Gaskin. Thank you again for the informative tutorials. I found that during the CLF, if we delete Joy_6, it will help a lot. Actually, I my view it's better to delete Joy_6 than Joy_7 because that when I check the output - standardize... Joy_6's loading is the least (maybe 0.28) and Joy_7's is slight greater. So the question is coming, what should we do when we get a better result after deleting Joy_6? That means that the video should be a slight changed after 13:23 ( and your cheer should be heard, I literally like your glad voice).
Hello James, First thank you for the great work and continued updates. This is a fantastic resource.
Second, here are my problems and questions. I have two measurements of the same model (one about the values of individuals the other about the values of an organization, using the same factors and items). The model has 6 reflective factors and 3 items per factor. Harman's single factor test revealed that the second measurement had CMF issues (67% of variance explained :( ) . Unfortunately, the CLFs won't run (for either measurements). I have 201 respondents and no missing data. The first measurement was pretty "clean" with Harman's test explaining 37% of variance, high loadings between items and factors (.76 and up) and covariances between factors are all under 55%, model fit is good. For this first measurement, CFL won't even run (I used your plugins for the Pattern matrix import and the CFL), even if I allow for 500 iterations. For the second measurement, I can run the model with CFL and it finds a solution after 101 iterations. The standardized estimates from the CLF are ridiculously high (1.0 and up) and there aren't any loadings between factors and items or between error items and items. (??)
Is my sample too small? Am I probably missing something? Would it be acceptable to combine the measurements in order to get 402 respondents and then test for invariance and CFL?
Note: I use SPSS24 and AMOS 24.
Sorry for the long post. Best, Nico
Sometimes the CLF just breaks the model. This is a known issue. (I can't remember the reference right now...) When the CLF is more trouble than good, it is okay to remove it and move forward with the possible limitation of method bias.
Thank you James! I was struggling with the decision to move forward. Hopefully, other aspects of the research will make up for this flaw :) Keep up the great work. Nico
Thank you so much for your amazing tutorials!!!
Thank you, James, for all your sharing about CFA and SEM!!!
I have a query about the zero-constrained approach for detecting common method variance.
It seems that the approach used in the video tutorial is more like the ULMC technique discussed in Richardson et al. (2009) (the CLF seems to correspond to the unmeasured method construct in the Trait/Method Model in Fig. 3B of the paper). But in the section of "Advice to Authors", Richard et al. (2009) "do not recommend using the ULMC approaches" (p.794) .... Did i misunderstand anything here ....? would you mind advising? Many thanks!
Hello, I have downloaded the Excel StatTools but there was no column for "past Variance table " only "Paste Correlations table in A2 and Paste Standardized Regression Weights table in F2" were appeared in validity master option .
ah yes, the tool I use in this one has already been updated again. I have another video on it in this playlist. Here it is: ruclips.net/video/8Sumk785Fmc/видео.html
Hi, james... Great job ! Even I am finding the same issue. No column for "Past Variance table". Thanks for the redirecting link of the data.
Hello James, I appreciate that you upload these tutorials on RUclips. Actually I have a question: After running common method bias test, if the results show that there is no common method bias, shall I remove the common latent factor and then impute all other factors to generate the new scores?
+Yang Feng Yes.
Hello again, Doctor James Gaskin.
I downloaded the Stats Tools Package.xlsm, copied 'n pasted correlations and Std.regresstion weights, and then gently pressed on the big red button.
Excel said "Run-time error 6: Overflow", and Microsoft said that because my data is "is too large to be represented within the range of values allowed for that type of variable".
I checked the 'Caveats and Assumptions', I think my data is met. Is there a way to deal with this error?
* Debug arrow points at 'CR = SL3 / (SL3 + sumErr)'
And, I tried 'MasterValidity' the AMOS plug-in to my data. It suggested me to keep removing observed variables, but CR of all factors still lower than 0.70 and there's no variable left to remove. In this case, is there a way to cure this problem? However, I tried CLF method and the diff.P was .999.
You can email me your excel file before pressing the button. I'll take a look. Google James Gaskin BYU to find my email address.
Hello, Prof. Gaskin. Thank you so much for the tutorials. I learned so much from you, more than my supervisor. I have a question, how to set or choose a mark variable?
A marker variable for method bias is usually chosen theoretically based on what we think might influence all variables in the study. Social desirability bias is the most common and the only validated marker variable that I'm aware of.
James
you are unique. thank you for all knowledge and video that you shared.
I have a suggestion. if it's possible, for any video notice reference. plz do not misunderstandings. I sure you did right, but you know in the article any sentence you write, have to writ refrence..
in this video i need reference of CMB method with CLF to notice in my thesis..
thank you James. thank you fo your being. :)
Thank you for the suggestion. References are definitely my weak point. As for this zero-constraint approach, the best reference is probably: Richardson, Hettie A., Marcia J. Simmering, and Michael C. Sturman. "A tale of three perspectives: Examining post hoc statistical techniques for detection and correction of common method variance." Organizational Research Methods (2009).
Here is a link to a list of references: statwiki.kolobkreations.com/index.php?title=References
yeah.
thank you professor James.
@@davoddavodnia7270 We have used this approach here:
Serrano Archimi, C., Reynaud, E., Yasin, H.M. and Bhatti, Z.A. (2018), “How perceived corporate social responsibility affects employee cynicism: the mediating role of organizational trust”, Journal of Business Ethics, Vol. 151 No. 4, pp. 907-921.
@@Gaskination If someone wants to cite this, we have used this approach in our paper published in Journal of Business Ethics, You can cite the following:
Serrano Archimi, C., Reynaud, E., Yasin, H.M. and Bhatti, Z.A. (2018), “How perceived corporate social responsibility affects employee cynicism: the mediating role of organizational trust”, Journal of Business Ethics, Vol. 151 No. 4, pp. 907-921.
@@xeeshan83 Thanks! I will add this to the statwiki
Prof, I got the error as: "When performing data imputation, Amos uses an iteration limit of 500. This limit was reached without successfully fitting the model. Imputation cannot be performed."
What might be the reason?
Any help will be highly appreciated
Try this: ruclips.net/video/B7YOv7hSohY/видео.html
Thank you SOOO much for these amazing resources! I recently downloaded your (also phenomenal) Excel Stats Package, but the Validity Master tab does not have the section for pasting the Variance table.
ruclips.net/video/8Sumk785Fmc/видео.htmlsi=viMIzRDfp28r5p39
Hi James. Thanks for your video! I have two questions. Can I use your method to test common method bias using Amos if one of my constructs has indicators that are dichotomous? May I know if the CLF connector plugin is still available? It seems I could not find it on wikista or your dropbox. Many thanks!
Hello, Doctor James Gaskin.
I do not know how to draw a single indicator variable in to AMOS.
Although I set a factor loading fixed to 1, but AMOS does not allow me to run my model.
Don't connect it to a latent variable. Just include it as its own variable (not with a latent factor).
Hi James,
Thanks for the valuable tutorials. They are extremely helpful. I have one question: is there any Information Systems literature to support imputing the factor scores? Most everything I can find says to not impute. It seems like the degrees of freedom would significantly change from the CFA to the structural analysis. Do you have a IS paper I can reference that applies this strategy?
Thanks for the help!
I also prefer to not impute. It is best to use latent models whenever possible. They do result in many more degrees of freedom than a model built with factor scores. The only case I can think of for using factor scores instead would be if you needed to do interactions (which get crazy-complex with latent models).
Hi James Gaskin, I found your videos are very helpful and thanks for sharing! I have discriminant validity issues with my PhD data and found many factors are highly correlated with >.70 values. I refer to the cut-off point value of .85 (Kline 2011), but I saw in this video that you considered .70 as highly correlated. Could you kindly suggest me the literature that supports this argument? Also, is it acceptable to delete constructs/factors that are highly correlated?
I recommend going with Kline, and then performing an HTMT analysis (ruclips.net/video/Mrr9rUWIzTE/видео.html) and fornell larcker criterion analysis (same video for both).
Hi James. Thanks for the great videos. I have a question. Should we check the reliability and validity after CLF?
I'm not sure what the literature has to say on the matter. I usually don't, although it would be good to ensure the CLF hasn't completely destroyed your factors.
Hello James! Thank you for your valuable videos. I'm confused in one thing that when their is common method bias present? If No comes, it means no difference between two models, and therefore no CMB, and if yes comes it means their is a difference between both models and CMB is present? Or otherwise? Please guide me because when I read your suggested reference article using ULMC technique which suggested that if two models are significantly different there is evidence for CMB!
No means not invariant (not not different). I know this is confusing. When it says NO, that means the models are different and so there is CMB. Yes means they are invariant (not different), and so there is no CMB.
+James Gaskin thanks a lot! Stay in touch ;p
Hi James, thank you for the SEM series, I have learnt so much from them. I am wondering if Data imputation in this video and Parceling technique have anything similar. To me, they are quite similar in a way that they both aggregate items and using those aggregates as indicators of latent constructs. Can you please tell the difference and similarity of those methods? The second issue is I cannot find Estimands file from Dropbox folders in your website anymore. How could I possibly have the Estimands file to test mediation? Thank you very much!
Parceling usually means creating averages across different dimensions of a higher order factor, whereas these composite factor scores are simply weighted averages for each latent variable. As for the estimands, they are available on the statwiki homepage via a link on the left navigation pane called "plugins and estimands".
Hi dear Gaskin. Thanks for your great videos!
I have a question about common method bias. In the video, you have shown that after analyzing CFA with a common latent factor, you reached the conclusion that there is a CMB and you imputed data with CLF in CFA model. I have used your method and found that I have a problem of CMB, as you have shown in the video. I don't know whether the CFA model with CMB is publishable in journals or not? Is the SEM model followed by imputed data of this CFA model reliable for analyzing or not?
It is fine if the factors do not break in the process. If the factors break (i.e., their indicator loadings change drastically), then try shifting around the parameter constraints on that factor until the loadings return to normal. Then if you still need to impute with the CMB included, the model is valid because it is controlling for the effect of CMB.
@@Gaskination
Thanks for your helpful response.
Could you please introduce me the reference allows us to impute data with CLF in CFA model which is valid for using in SEM model?
All published papers I have studied reported that fit indices with CLF are worse than the ones without CLF, so reported that "common method bias was not a major problem". Based on these published papers, I thought that a CFA model with common method bias and the following related SEM could not be publishable.
I have common method bias in my data based on the analysis in your RUclips film, and I really need a reference to justify the presence of common method bias in my data, and imputation with CLF.
I would really appreciate it if you let me know the reference.
@@fatemehnili4723 I recommend using the approach utilized by the articles in your discipline. If they all use the model fit comparison approach, then I recommend that approach. If that approach fails, then you could use this approach: ruclips.net/video/abzt5zTkCxk/видео.html
Dear Mr. Gaskin,
thank you very much for your helpful videos. I do have a quick question. I have a model with multiple latent factors (A, B, and C). Whereas A and B are indicated by four variables each, the factor C is only indicated by two variables. When I compute the CLF as advised in your video, the model won't run anymore. When I add a third variable to C, it all works fine again.
My question is: Is there a constraint I am not aware of regarding the number of items needed for an indicator in order to run a model with the CLF included?
Thank you very much for your help.
The CLF usually makes things unstable. Having two indicators for a latent factor can also make it unstable. These both together will very likely break the model. The CLF is also mostly good for complex models, rather than small ones. Here is my most recent video on this topic: ruclips.net/video/abzt5zTkCxk/видео.html
Dear Mr. Gaskin,
thank you very much for your fast reply. I really enjoy your work and you are helping a lot of us PhD students.
Hi James,
Thank you for the tutorial videos for SEM series using AMOS. I am conducting CFA using AMOS with the data set of 593 and five latent variables. The CFA model fit is good (Cmin/DF= 2.28; p=0.00; CFI=0.935; GFI=0.968; AGFI=0.947; RMSEA=0.047; PCLOSE=0.69) but when checking for the Composite Reliability (CR) and Average Variance Extracted (AVE), I am getting the values less than 0.7 and 0.5 respectively (see below the summary). In the other hand, Discriminant Validity is looking good. Would you please suggest whether I can go ahead with these results (any reference)? If not, is there any way to improve this?
I have already performed the EFA and removed some indicators to make the pattern matrix better. I tried hard but unable to make total cumulative variance explained more than 45% and KMO value 0.636.
In addition to this, I have a dependent variable which is in the form of bivariate nominal variable i.e. not on a Linkert scale. Can I add this variable while solving structural model after completing CFA?
Thank you in advance.
CR AVE MSV MaxR(H) A B C D E
A 0.621 0.471 0.153 0.749 0.687
B 0.733 0.484 0.056 0.766 0.230 0.696
C 0.592 0.293 0.065 0.715 0.159 -0.024 0.541
D 0.596 0.467 0.065 0.812 0.060 0.137 0.255 0.683
E 0.645 0.489 0.153 0.734 0.391 0.236 0.117 -0.063 0.699
When I see CR and AVE this low, it makes me wonder if the factors are truly reflective, or are actually formative. As for adding the DV in after the CFA, yes, you can do that.
Hi James,
Thank you for these videos. Is there a way to find the significance level of the correlations among each variable produced in the ValidityMaster table? Thanks for the help and for sharing all of the videos with everyone.
Chris
I suppose you could use the p-values created for the covariance matrix. I think I'll edit the plugin to include this in the future. Thanks!
Dear James.
Thanks for all your help till now. Can you please provide references for MSV
No one uses MSV, so you can probably just not put it in your report. The Fornell-larcker criterion is sufficient (square root of AVE greater than inter-factor correlations.
Hi James, thank you very much for the excellent videos. My problem is that after including the CLF for the Common Method Bias assessment, I get multiple heywood cases with standardized regression estimates >1 that will not go away with the techniques you describe in your other video. Is this a common occurrence or is there another way of approaching this in this case? I have several 2nd and 3rd order constructs - could that be the root problem?
That higher order model might be the problem, especially if those factors have only 2 or 3 dimensions. Also make sure your factors are reflective, rather than formative.
Thank you for all these videos! I cannot really tell how grateful I am for these! But when I download this Excel tool from Wiki my file doesn't have this variance table in Validity master page :( How can I get this updated version which you also use in the video?
Here's an update: ruclips.net/video/8Sumk785Fmc/видео.html
Can you share the best reference for the Modification Indices to Improve Model Fit as it is mostly discouraged by the researchers to perform this action. Do you have any solid reason if we are making this "Modification Indices" in model fit improvement. Thanks
This video is six years old and I have also shifted to agree with the discouragement of adding modifications to a measurement model. Instead, it is better (usually) to omit redundant items. Here is a good reference against adding covariances between errors: Hermida, R. 2015. "The Problem of Allowing Correlated Errors in Structural Equation Modeling: Concerns and Considerations," Computational Methods in Social Sciences (3:1), p. 5.
@@Gaskination Thanks for the clarification. Your videos helped a lot.
Hi James. Quick question. Is it an issue to report AVE and CR for each group? I am thinking to create a table and compare AVE and CR between the "ALL" measurement model with each group measurement models. Thanks :)
If the groups are invariant, then the AVE and CR should not change substantially across groups, and therefore only needs to be reported for all the data at once.
Cool. Thanks James. Yeah I got AVE=.76; CR=.95 (no groups measurement model) and AVE=.78; CR=.94 (first group measurement model) and AVE=.74; CR=.95 (Second group model). It's not a big change. I just want to guarantee the readers that AVE and CR in all types of measurement model is very similar :)
Dear James, first of all - thanks so much for your videos! I am using them myself, and recommend as an reference to my students! I have a question about imputation that you discuss at the very end of this video. When I run it, I get a message from Amos that it uses "an iteration limit of 500. This limit was reached without successfully fitting the model. Imputation cannot be performed". Did you ever face this? if yes, what would be your recommendation how to deal with it?
I am facing the same problem. How did you fix it?
Yes, even I did face it. Did you fix it?
Dear Prof.
I watched your youtube video and then downloaded the Stats Tools Package.xlsm, copied and pasted correlations as well as Std. regression weights in excel file, then pressed on the red button. However, It got problem "run time error 1004: application defined or objects defined error", I check many times and do as your instruction but I can not solve the problem. So is there any method to deal with this problem?
Thank you so much
You can instead use the plugin: ruclips.net/video/JqySgMU_qMQ/видео.html
Hi James, thanks for great videos! following your instruction, i got this message and i don't know what am i supposed to do;"The model is probably unidentified. In order to achieve identifiability, it will probably be necessary to impose 1 additional constraint."
This happens when you are missing a constrained indicator regression weight for one of the factors. Check to make sure one indicator regression weight is constrained to 1 for every latent factor. If you are using a 2nd order factor, make sure one of the paths between 1st and 2nd order is constrained.
@@Gaskination Hi James, I have the same problem but there is no "1" missing nowhere. What can I do?
@@CarinanaLaa This should not occur unless there is a missing constraint. If you have contained the paths on a factor to be equal, then you'll need to put the constraint on the variance of the factor. If that is not the issue, then perhaps you have a duplicate variable that is overlayed and therefore seemingly invisible.
Hello James, thanks alot for your videoes, i think they are the best on youtube for this subject, well done! - I've a question, the last part where we are supposed to impute data i get an error because i've a -.11 between 2 items. How do i fix that? - also i have an item with 1.02 (harley case), how to fix errors like that?
+Eric Frederiksen I'm assuming you mean there is a negative error variance, not a negative correlation. If it is a negative correlation, that is fine. As for a heywood case, here is a video: ruclips.net/video/Vx24KFf-rAo/видео.html
+James Gaskin Thanks Sir!
Hi James. I really appreciate all your videos on SEM. Thanks for that :) I do have a question regarding the C.R.: When I conduct the CFA, all CRs are just fine (I use 2nd order constructs). But when I move on to the path model, some C.R.s are not reaching significance anymore. Can you help me with this problem? I did not impute the factor scores, instead I just added paths between my latent variables.
Do you mean the critical ratios or the composite reliabilities? If reliability, then it could be due to the weak regressions between that factor and other factors. It is not a problem for reliability at that point.
@@Gaskination Hi James! I was talking about the Critical Ratios, sorry about the confusion! So in my measurement model all Critical Ratios are finde but when I move on to the path model one first order Indicator of my 2nd order construct suddenly has a Critical Ratio under 1,96. I am able to fix the problem by specifiying the construct as a first order construct. But I don't understand why this is happening.
@@marysmith2153 It's just because you are relating it to other factors that may be strongly pulling the centroid of correlations from indicators away from the original center.
Dear Mr. Gaskin, many thanks for your awesome videos - they are purely great and support my research!
I got one question concerning the CMB: I had problems with CMB during a factor analysis and I solved them with your help with the CLF. Surprisingly, my global fit increased which is perfect! Afterwards, I imputed my factor score of the resulting 2nd order factor - am I now able to work with this score in SPSS (I want to calculate a multiple regression with its help and used AMOS to test for model fit, reliability and validity (your 6-step-approach))?
Yes, you can use these factor scores in regressions and path modeling. You do not need to use the imputed CLF factor score however, as its effect is already accounted for when imputing the other factors.
Dear Mr. Gaskin, thank you very much for your answer! Just one more (short) question: Can I report this method by stating that I used the CLF-Method (cite your source which you posted quite often already under this video) and thereby improved my model fit? Since this is, besides solving the problem of CMB, the major contribution of the method, it significantly improved my fit measures.
Again, thank you so much in advance! Your videos are awesome!
That should be fine. The reason fit improves with the addition of the CLF is because you are now accounting for a lot of variance that was previously not being accounted for.
If someone wants to cite this, we have used this approach in our paper published in Journal of Business Ethics, You can cite the following:
Serrano Archimi, C., Reynaud, E., Yasin, H.M. and Bhatti, Z.A. (2018), “How perceived corporate social responsibility affects employee cynicism: the mediating role of organizational trust”, Journal of Business Ethics, Vol. 151 No. 4, pp. 907-921.
hi James, thanks for the excellent works. The excel stattools that i got is 3/5/2016 version, is this the latest version?
yes, that is the latest version.
Dr. Gaskin, Thank you so much for sharing your SEM Lecture series.
I have a question regarding CFA with control variables.
It seems that CFA is tested without control variables but SEM runs with control variables (e.g., job performance). Why dose not include control variables in CFA?
Thank you in advance!
CFA only includes reflective latent factors. If your controls are reflective and latent, then you should go ahead and include them in the CFA to ensure they are discriminant from your variables of interest.
James Gaskin I respect you and thank you so much for your clear answer.
Hello James, I am unable to compute the CR, AVE MaxR(h) for some variables using your excel tool as it returns as error message stating that 'division by zero'. I can calculate the CR & AVE using the formulas provided and they return all positive logical values but it doesnt when using the excel tool
hi James, i found the error. I have two factors which are loading exactly at 1, which is causing the division by zero error. Can you help me to fix it. I tried to resolve it with Heywood case solution but it just threw off the entire model and I cannot remove these factors as they measure the Dependent Variable. Shall i just assume a factor loading being very close to 1 like 0.9999 but not exactly 1.
Hello Dr. Thank a lot for the video. I do not know why CLF connector plugin is not working well for me (Amos 23). However, when I I draw manually CLF latent variable and connect it to all Items by drawing paths I can get results. Thanks for your help.
Dear Dr. Gaskin,
first of all, thank you for your video, I am a beginner in using amos and statistic, and your video really help me a lot to learn amos
I have some trouble during cmb, my path diagram (unconstrained) can't be calculated due to iteration limit reached.
After looking the view note, I found that some of indicators (from one latent variable) have very high standard error and weird estimate number (e.g estimate= -92.416 s.e= 1276.899). Do you have any advice for my problem? Thank you
Here is a video to help: ruclips.net/video/B7YOv7hSohY/видео.html
Hi, Thanks for this HELPFUL series and for all the knowledge you provided us with. I just want to ask how to install a plugin such as the common latent factor (CLF) in AMOS 23 . Thanks in advance
Hi Dr. Gaskin! I hope you do well and continue your great job! I have a model with 5 IV and one categorical(nominal) DV. I have run an EFA/CFA with the 5 IVs only and finally produced imputed factor scores. Is this OK to use these scores in a multinomial logistic regression to predict the DV? I read about a certain software that handles models with categorical DVs but I do not have that option at least for the time being. What would you suggest in such situation?
Yes, you can do this. SPSS offers this functionality.
Could I use path analysis with nominal DV? Thank you.
@@ysitar Yes, but only if it is binary. Not multinomial.
Hi! Thank you for the tutorial videos and resources. I downloaded the Excel StatTool from your webpage, but in Validity Master I can only insert two tables (correlations and SRW). Where could I get the latest version that you use in this video? Thank you for making statistical analysis understandable and easier!
Now I saw in the description of the video the link to the correction video. :)
@@maiki7444 I am still in the same problem with you. Could you tell me how to do it?
Dear James, during CMB test i get a error message stating that the model is probably unidentified. in order to achieve identifiability needs to impose 1 additional constraint. i have checked for the constraints various times. everything seems to be constrained?? please guide.
+SANDEEP MUWAL Is there a variable floating around somewhere outside the current view? Or, check the CMB if it is constrained to 1 for its variance.
Professor Gaskin, thank you again for your videos, I'd like to ask why you don't include moderator into the analysis at this stage?
The only reason to exclude the moderator during the CFA is if the moderator is not latent.
@@Gaskination Dear Dr. Gaskin, I have a query that builds on good sir's (Юлия Музыченко) question above - Had the moderator been a latent factor, should we have included that as well in CFA and subsequent CMB analysis? - Sincerely, Tushar
@@tusharprabhakar9653 Yes. Good question. For the most robust and valid results to a measurement model (EFA/CFA), use all reflective latent factors, even if they are not IV or DV.
@@Gaskination Thank you very much, sir! I appreciate your generosity.
Dear James n his followers
Greetings!!
Please advice me on the below mentioned query/doubt -
What should be done when Harmans one factor test during EFA using SPSS predicts no CMB while chi-square test during CFA turns out to be significant hence implying CMB..
How should this be reported esp when I retained CLF for further analysis of SEM (not really in a position to change that now)
n as always thankful for these series and videos..
The CFA test is considered more rigorous. So, as long as adding the CLF did not break your factors, then I would rely on that test.
Thankyou so much for the prompt response and clarifying my doubt
Hi James. Firstly, thanks ever so much for your fantastic videos and helpful tools.
I have a problem with my unconstrained model in AMOS that I wandered if you can help me out with. When I run a CMB/CLF on four of my latent variables, one latent construct doesn't produce any standardised values it does produce unstandardised values. When the items are fully constrained all latent variables produce unstandardised and standardised values.
I then move on to data imputation. When I impute my data an error message saying "A sample of parameter values were inadmissable".
Would you know how I can overcome this problem?
Thanks so much!
The CLF often breaks the model. You can try some of the strategies in this video: ruclips.net/video/B7YOv7hSohY/видео.html
If that doesn't help, you might just remove the CLF and report that it broke your model.
Hi James thanks for the tutorials. I v a question, when I download the Stat tool package from your statwiki page, it still shows only the correlation table and standardize regression weights. How do I get an updated one with a function to paste variance. Thanks
The variances column was a mistake. The one that is on the statwiki is the current and correct one.
Hello James - I have a questions. How does the approach of testing common method bias differ from a second order CFA?
CLF is linked to all observed items. 2nd order factor is linked only to its first order factors.
Hey James, thanks a lot for the assistance, I noticed the the latest excel stattools file still does not have 'paste variances table in k2' column, in the ValidityMaster tab.
Correct. The variances column was a mistake, as explained here: ruclips.net/video/8Sumk785Fmc/видео.html
Dear james i want to know that is it necessary to check discriminant validity among first order factors while applying second order CFA?
Sorry I didn't see this earlier. RUclips didn't notify me for some reason. If the 2nd order factor is reflective, then we expect the first order dimensions to be fairly tightly correlated. My guess is they will not have discriminant validity.
Hi James, thank you for this informative video. I completed all the steps but right at the end when I did the data invitation I had a messaged saying the following ‘a sample of parameter values was inadmissable’. Followed by ‘1 parameter samples were inadmissible. 0 completed data sets have been created’. Do you have any advice on how to solve this problem please? Thank you :)
Check the notes for the model in the output window. Usually it will point you to the problem. This error is often associated with negative error variance. In such a case, you would just need to constrain that error variance to a small positive number, like 0.05
@@Gaskination Thank you for the speedy reply. I tried this but unfortunately the issue with data imputation still persists with the same message. Thank you anyways, your videos are very helpful.
@@elliemoore8896 Data imputation won't run if the model won't run. Make sure to address errors in the model before trying to impute.
@@Gaskination Thanks James. The model does run. It runs successfully when the CLF is unconstrained and fully constrained. It was just the final stage of trying to impute the data where I fell into this error. Can I proceed without data imputation? Perhaps I can manually create variables for each factor in SPSS instead?
Hello Prof. James Gaskin
Thanks a lot for your helpful videos. I am writing my paper and I read the reference you wrote below (Richardson et al., 2009). It seems authors do not recommend to use latent method construct.Is there any other newer method to measure common method bias? Thanks.
No one strongly recommends the approach. There are no great approaches to extracting CMB. There are some references that might help on the references page of the statwiki: statwiki.kolobkreations.com/index.php?title=References
Search on that page for the word "bias". There are three results.
Here is another: Podsakoff, Philip M., Scott B. MacKenzie, and Nathan P. Podsakoff. "Sources of method bias in social science research and recommendations on how to control it." Annual review of psychology 63 (2012): 539-569.
Thank you very much!
Hi James, thank you very much for such an insightful video on CFA and SEM. However, I would like to inform you that while trying to conduct reliability and validity test using the excelStatTool provided in statwiki, responds about corruption in macro file hence no result is derived. I have attempted many times and have even checked for the conditions you have stated in the excel sheet. The MACRO seems to have gone corrupted.
Thank you for reporting this. You can either try to download a new file, or you can try this plugin: ruclips.net/video/JqySgMU_qMQ/видео.html
With James permission, have you tried to enable excel files from internet? Excel tends to block them, specially if they have macros.
yup all done
Hi James, thank you for your very helpful videos. I was wondering whether you have any references for the common latent factor test? In my questionnaire, I have included only one item as a marker variable not three, and my model in AMOS does not run. I would really appreciate any tips. Thank you in advance.
Also I cannot download the CLF plug-in, every time I try, secure download manager opens and message is displayed that “the navigation to website was cancelled.”
A single item doesn't work well as a marker variable. To include it in AMOS you might try sticking it in as just an observed variable (not attached to a latent factor). As for the plugin, you might have to try using Chrome. Microsoft Edge and Internet Explorer will probably forbid the download.
Excellent, I found your paper "taking 'fun and games' seriously: Proposing the Hedonic-Motivation System Adaptation Model (HMSAM)".
Thank you for all your superb videos and of course for your responsiveness to all the questions! Very much appreciated!
Dear James, I am having few problems, which I can't resolve. Firstly, when I try to do 'Common method bias', I find the up-arrow not highlighted and I couldn't get the items and constructs weights or loading. Secondly, when I try to impute, I get this message 'an error occurred while attempting to fit the model Amos imputation cannot be performed'. Could you kindly help please.
Watch this video for clues: ruclips.net/video/B7YOv7hSohY/видео.html
Thanks James, I resolved the issue, what I did is that I removed '1' from each of the 1st item and relocated on the construct and it worked then. I will watch the video and will leave my comment. Thanks, you are very, very kind. Loads of respect and love :-)
I have watched your video now and I will follow your recommended steps too. Once again many thanks for your efforts and thanks for such a great help!!!
Hi James
Thanks for the awesome tutorials. I was doing CFA analysis on my data before that I did EFA and I got a clean pattern matrix with all factors loading above 0.7, I used AMOS 26 with your plugin to to do CFA I copied my pattern matrix and run the model, after I copy the correlation table and standardized regression weights table I get the same problem you encountered in this tutorial, one of my constructs square root of AVE is less than absolute value of the correlations with another factor, I followed the same solution you did, I removed one item from my construct and after that the construct remained with only one item, I saved the changes and run the model but when I see the model fit after this I dont see the standardized regression weights table anymore.
Any help will be highly appreciated
The CFA requires all latent factors have at least two items. If the factor doesn't have a strong AVE, consider whether it might be formative rather than reflective (here is video explaining that: ruclips.net/video/gw0xvvJw-AM/видео.html).
Hi, On running the zero constrained model I'am getting this on output "The model is probably unidentified. In order to achieve identifiability, it will probably be necessary to impose 1 additional constraint.
" What shall i do next?, I have checked that regression weight is constrained to one for every latent factor.
This only occurs when a path constraint is missing from a latent factor. Make sure if you have multiple groups that the path constraint is there for both groups.
Thanks so much for your very useful video. I have some questions. My model have less reliability and validity on each constructs (just 1-2 that is ok from 6 construct). I try many ways to improve the reliability and validity but it doesn’t work, although the other values show good model fit (e.g. RMSEA, PCLOSE, CFI, GFI, CMIN/DF). Can I still continue to the next process; impute factor score and SEM? Do it will have any problem with less reliability and validity? Thank you.
It is very important to reach these thresholds of validity and reliability. Otherwise you cannot have confidence in your findings. Usually if you can achieve a good solution in the EFA, then the validities in the CFA are easy.
James Gaskin Thanks so much for your reply :)
Hello Dr. Gaskin, thank you for these great tutorials. I was trying to do the common method bias but after using the plugin, when I try to run the model it just do not run. I just can't get over this... can you help me please?
It is very common for the CLF to break a model. This happens most often when there is low sample size, or strong correlations between factors. Here is a video that might help you resolve it: ruclips.net/video/B7YOv7hSohY/видео.html
hi James, thanks for sharing this great video. What is the reference for using this Zero-constrained approach please? thanks
Sorry for the delayed response. Here is the reference (also available on the references page of the statwiki):
Richardson, Hettie A., Marcia J. Simmering, and Michael C. Sturman. "A tale of three perspectives: Examining post hoc statistical techniques for detection and correction of common method variance." Organizational Research Methods (2009).
many thanks James.
Hi, could you please explain the error: "the AVE for XXX is less than the MSV" & "the AVE for Control is less than 0.50." and How to fix them, Thank you
The MSV greater than AVE means that there is probably a discriminant validity issue. However, this can probably be fixed by increasing the AVE (which is also low). The way to fix AVE is to remove the lowest loading item. If that doesn't work, you can try figuring out which item should be removed by following this video: ruclips.net/video/xVl6Fg2A9GA/видео.html
Hello Prof. James Gaskin,
do you have any video instructions how to do a multivariate normality test?
I have to assure that normal distribution is given for all my varibles.
Thanks in advance
Multivariate normality is tricky because it will always exist, even if you remove cases that violate normality, others will take their place. Here is my best video on the topic though: ruclips.net/video/J2EkjIeK-PE/видео.html
Pay attention to Cook's distance.
Your videos and your Excel Stat Tools are excellent. I'm using them for assessing construct validity in my model, but I have a doubt. As I have some correlated error terms in my model, I'm wondering if the formula you use to calculate CR already deals with that. Because when I apply the formula you present I don't have the same result as the Excel Tool. Can you help me with this? Thank you!
+Elsa Pedroso The new tool accounts for error variances. If the errors are covaried, it might have a strange inflation effect.
+James Gaskin But the results for CR shouldn´t be the same for the two Excel Stat Tools files (the old and the recent one), for all the constructs which don't have correlated errors? My model has four constructs, only two of them have correlated error terms. When using both Excel Stat Tools files, CR values are also different for constructs which don't have correlated errors.
+Elsa Pedroso I fixed it today. There was a problem... Sorry about that. Here is the new video: ruclips.net/video/8Sumk785Fmc/видео.html
+James Gaskin Thank you! Now I'm wondering if the new tool deals with the error terms correlations to calculate CR. It doesn't seem to...
+Elsa Pedroso The new tool uses proportional error instead of exact error variance. As for the error correlations, the effect of those are reflected already in the estimates and correlations.
Hi James. Appreciate the videos. Just wondering whether you have any workarounds for the CLF plug in, I'm unable to install any plugin onto University computers and want to know if there is any way around doing so to get this working. Thanks
You can always do it manually (i.e., draw the lines one at a time). Not much fun, but I don't know how to bypass the university security protocols.
Dear Dr. James, Warm greetings from India. I continue to learn a lot from you. I was trying to use the master validity tool and when I place the correlations and standardized regression and click, I get the message "cannot run the macro....the macro may not be available in this workbook or all macros may be disabled." Can ou please guide? I have downloaded the excel file from Wiki Stats. Regards, Vikas Rai Bhatnagar.
You need to enable macros. If you don't see the button at the top that says "enable macros", google it. Also, you could use the plugin instead: ruclips.net/video/JqySgMU_qMQ/видео.html
Can I fix the error term variance to 1. What is the criteria?
Is the fixed regression weight to one observable item should be removed while fixing variance of error term to 1 or we will additionally constraint the error term while having with one of the regression weight of observeable item fixed to 1?
I mean that we can only constraints one at a time like 'variance or regression weight' for "latent variable" and same is the case with the "error term". What if we constraint the 'latent variable to 1 along with error term constraint to 1. I am confused. Can you elaborate it further?
Yes. Just doubleclick the error term and go to the Parameters tab to constrain the variance to 1. The one observed item with a constraint is not redundant with the error term variance constraint. You'll still need the constraint on the regression weight for one observed variable.
Thanks Prof James for the great video. My data had significant CMB. May I ask for the reference of the imputation method? Thanks in advance.
I'm not sure about a citation for it. I'm pretty bad with citations... Here are some potentially helpful references: statwiki.kolobkreations.com/index.php?title=References#Confirmatory_Factor_Analysis
I'm not sure about a citation for it. I'm pretty bad with citations... Here are some potentially helpful references: statwiki.kolobkreations.com/index.php?title=References#Confirmatory_Factor_Analysis
Thank you so much for your prompt reply. I am going to read those for more information.
When I try to download the excel the old version without the option to post the variance table is downloaded :/
ruclips.net/video/8Sumk785Fmc/видео.htmlsi=x_ozacMpXQoQZ62w
Hi, i'm using Amos 22, probably I've tried every possible way you mentioned in statwiki and in here but i couldn't instil PLS plugin in amos it doesnt show up in plugin list.
Sorry. I'm not sure why. Others have found that if they don't unblock it (in file properties window) then it won't show up. And some have found problems if they don't run amos as administrator.
Thanks Sir for this wonderfus turos. When I lanch CLF plugin, i get this message "iteration 49" itiration limit reached . I get the parameters (khiI² and degree of freedom for both constrained and unconstrained models, i also get outputs sheet but output path (red arrow) always still greay (disabled). I have 8 latent variables and 27 items ... may be the plugin doesn't support complexe models? can anyone help please? thks. Is it possible to do CBM manually? by drawing all arrows and values 0, 1?
I fix the probleme. check out this: watch?v=B7YOv7hSohY
Sir Where how can I download the new version? on my web, I only can see the old verson
There was an error in the calculation of the CR. Here is the new video: ruclips.net/video/8Sumk785Fmc/видео.html
@@Gaskination Thanks sir
But I would like to know the verson to calculate ASV
I think I watch it from one of your vedeos
But I can not remember
Sorry about my broken English
@@lsummer7060 Not many people use ASV, so I removed it. ASV is the average squared correlation. So, to calculate it, square all correlations with that factor, and then average those squares.
Dear Dr. Gaskin,
Greetings, one of my construct CR>0.7, but MSV>AVE. Therefore, I faced discriminant validity problem. How to solve discriminant problem in my study? Thank you for your time. :)
Regards,
Chin
Sounds like it has a discriminant validity problem. In such a case, find which items are most strongly crossloading between the two factors (one way to do this is to do an EFA with just the items from those factors), and then remove items accordingly until you have discriminance.
Dear Dr. Gaskin,
Thank you so much. :) A lot of helpful video you shared.
Dear Dr. Gaskin, thank you for these helpful and interesting videos. I am performing SEM on LISREL, however when I run the program, it says 'model does not converge'. Can you please advise on this. Thanks
I've never used Lisrel, but that error means it is having trouble reducing error. This happens when you have small sample size and/or bad validities. If you can do an EFA in SPSS first, that will help.
Thank you for your reply. EFA was the first thing I did and my sample size is 20 times larger than the number of unknown parameters. Can you please suggest what can be done in case of bad validities? Or if it is having trouble reducing error, what can be done to resolve this issue? Thanks
If the EFA turned out well, then it is very odd that it won't run in the CFA. The only thing I can think of is if you have multidimensional constructs modeled as unidimensional, or if you have formative factors modeled as reflective. Otherwise, it is a bit of a mystery. If it were me, I would swap out each item one at a time to see if it is a specific item causing the problem.
Thank you for your response Dr. Gaskin. I shall be doing same, i.e., swap out each item one at a time and see if it work. You've been of great help indeed. Thanks very much.
Mr Gaskin, can you please inform me how to calculate Discriminant validity for a Uni-dimensional scale
Discriminant validity is not relevant to a single scale. Only convergent validity and reliability are needed.
Hiallo, Mr. Gaskin. If I want to use the excel table to check my validity and report my results, how do I cite the source? Thanks for the useful video!
Gaskin, J., (2016), "Name of tab", Stats Tools Package. statwiki.kolobkreations.com
Thanks a lot!
Hello James,
Can you share some reference papers related to this procedure of Common Method Bias. Moreover, more info. regarding imputing factors to overcome CMB is required. Do throw some light on the this concern. Would be off great help..
The zero-constraint approach is best supported by this article:
Richardson, Hettie A., Marcia J. Simmering, and Michael C. Sturman. "A tale of three perspectives: Examining post hoc statistical techniques for detection and correction of common method variance." Organizational Research Methods (2009).
As for imputing factor scores, this does not overcome CMB. Instead, it accounts for it so that you are using factor scores that adjust for method bias.
Hi, Thanks for this series. I just want to ask how to install a plugin such as the common latent factor (CLF) in AMOS 23 . Thanks in advance
Here is a wiki page explaining how: statwiki.kolobkreations.com/index.php?title=Plugins
Hi James, thank you for your videos , I am currently doing the CFA for my masters dissertation, but I can't seem to access the statswiki website to download the stats packages for excel and others, please whats the new website address ? please can i have a drive where i can download the files and plug-ins?
statwiki.gaskination.com/
@@Gaskination thank you very much
Prof, i can't run it at minute 7:14 when i tried to hit the up arrow. what goes wrong and what should i do? thank you very muchh
This can happen if the iteration limit is reached. Here is a video about resolving that: ruclips.net/video/B7YOv7hSohY/видео.html
Hi James, what to do if the measurement and structural fit indices are same? Thanks
That is not unexpected. If the degrees of freedom are the same, then the structural model will have the same fit as the measurement model.
@@Gaskination 🙏🏻🙏🏻🙏🏻
Dear Sir when I use CLF and run I did not get output. So no red up arrow. I can not check those chi square. do you know the reason.
This video may help: ruclips.net/video/B7YOv7hSohY/видео.html
Thank u soooooooooooooooooo much
Hello Doctor James, your tutorials on RUclips helped out me and my fellow PhD students alot in undertanding AMOS and SEM. Thank you for sharing and effort!
Are you plaing on doing some tutorials on second order latent factor models? Looks like they bring some additional problems during EFA, CFA and CMB.
Could you please comment on the error I keep geting from AMOS while I trying to do CMB as it is described in your tutorials:
"During the analysis of a bootstrap sample, an attempt was made to compute a standardized regression weight between two variables, one of whose estimated variances failed to be positive."?
Unchecking "Standardized estimates" does let it run but what the point (this was IMB website help desk solution). Changing bootfactor to something higher than 1 in properties make it possible to run the analyses but it gives same standardized weight for constrained and unconstrained models.
What might be the problem with model/CMB?
1. I have a few videos on 2nd order factors, although I'm not totally comprehensive on the topic.
ruclips.net/video/gw0xvvJw-AM/видео.html
ruclips.net/video/HBQPqj63Y7s/видео.html
2. Find the error that is negative (look at notes for the model) and then constrain it to some small positive number (like 0.001).
Dear Sir, Iwas download stats tools package on your your StatWiki, The Quation is, why I always fail wen using Validity Master the error is Run-Time Error 13 Type Mismatch. what should I do, please help me sir thx u
Check the box of caveats and assumptions listed below the button. If these are not the problem, then try the plugin instead: ruclips.net/video/ekICmx_qcWg/видео.html
Sir...when I run the excel, it gives run time error 9...subscript out of range. what can i do
check the orange box below the button to make sure you are not violating any assumptions. You can also try the plugin instead: ruclips.net/video/Mrr9rUWIzTE/видео.html
HI sir.thx for the wonderful videos.sir one of my estimate(loading) in confirmatory factor analysis is 0.07 while all others are greater than 0.7.but when i delete this item,my whole model fit disturbs.my results and model fit becomes adequate if i retain this item.is there any reference to retain such loading in order to get the desired results
It is very strange that this one item, that is not loading very well, should be the lynchpin. I'm not sure what might be causing that. Try moving the parameter constraint around to a different indicator to see if this helps. You can do this by double-clicking on the path from the indicator with the constraint, deleting it in the parameter tab (regression weight box) and then clicking on a different path to add a 1.00 in the regression weight box.
@@Gaskination thx alott sir
Sir my stat tool package is not working how can i calculate CR, AVE, MSV and ASV
Check the caveats and assumptions to make sure you are not violating any. If it still doesn't work, you might try this instead: ruclips.net/video/JqySgMU_qMQ/видео.html
it says "a sample of parameter value was inadmissible".... what does that mean?
I now deleted one construct..... n it worked....
Hi which construct did u delete and how did u consider deleting it?
Hey, I've been trying to access the Stat Wiki page but am running into some issues. It seems that the page is not working. I've tried on multiple computers now. Does anyone else have the same issue, if so, is there anything I could do to solve it? Thank you to James Gaskin for the amazing resources and tutorials!
Yes, it went down yesterday morning due to a PHP update. I've been packed with meetings and haven't had a moment to fix it. I plan to spend some time fixing it later today. Hopefully it will be back up and running before the end of my day.
@@Gaskination Thank you very much - the excel tool is a lifesaver! I've noticed that the validity master requires more than 2 latent variables to run. I only have 2. Is there a way to sort through this issue?
@@alieenamathew8595 The way around that is to use the plugin instead: ruclips.net/video/ekICmx_qcWg/видео.html Since the wiki is still down, here is the direct link to the plugins folder: drive.google.com/drive/folders/0B3T1TGdHG9aEbFg1eEpqOWtrR3c?usp=sharing
@@Gaskination Thank you!
Sir can we connect from item on one factor to item on another factor sir?
It is not good practice to do this. If the items are related, let them share variance through their factors, rather than directly.
Hello james.superb videos.can u help me.when i try to impute factor scores ,amos gives the a sample of parameter values was in admissable
Check the notes for the model in the output window. There is probably an error variance that is negative. You can fix this by constraining it to be some small positive number (like 0.05).
@@GaskinationThank You so much for Your videos! I do have the same problem, but cannot find a negative error variance. Do all factors have to be correlated to impute? If i delete one covariance arrow between two factors it is working... I really would appreciate Your help!
@@teresajames2191 Check the notes for the model for any other error notifications. If there are none, then check modification indices for any extreme values. If there are none, then check model fit and validities. If no problem there, then go ahead and remove the correlation.
Hi James, many thanks for all your efforts. I'm using AMOS25 and trying to use MasterValidity plugin. will it work with me??? because it gives this message (MasterValidity file contains syntax errors)!!! need your help, many thanks in advance
Make sure you’re using the one in the 24+ folder (rather than the one in the 23 and lower folder). It should work. Here is a troubleshooting guide that talks about syntax errors: statwiki.kolobkreations.com/index.php?title=Plugins#Troubleshooting
many thanks @@Gaskination
Sir, I ran the model for Data Imputation in the end. I got an error and data Imputation failed. I saw a note and it indicated that "covariance matrix is not positive definite". How can I resolve this issue? Thanks. Please reply fast in this case. I can share my data screen to look it for you if you allow me to in personal..Thanks
Make sure the model runs normally before trying to impute. If the model won't run, it won't impute. If you have a common latent factor included, try removing it, as it can sometimes break a model.
@@Gaskination Yeah, It worked fine by removing the Common Latent Factor. But how will I justify this CMB case as it was significant in my case as (Chi square>3.84 for Degree of freedom=1). It is significant. But failed to impute while incorporating Common Latent Factor.
@@abdulmoeed4661 Wow! So few degrees of freedom. That is uncommon for a CFA. You can just say that the CLF broke your model and wouldn't allow you to proceed. Therefore it was removed during factor score imputation.
@@Gaskination Thanks you very much..
Hello sir I want to know that why I can't run my model in Amos?please let me explain whole process of execution of programs so that I can understand my problem..
Not sure how I can help unless you give me more info. Do be as concise as possible though. Thanks!
James Gaskin okay..let me know why you took data from pattern matrix for cfa?what is reason behind it?
The pattern matrix is the factor solution for the EFA. We use this factor solution as input into the CFA because it is a valid way to assign items to factors. If we were to ignore the pattern matrix, we would very likely result in a poor CFA.
James Gaskin thanks again..I will try this info on my model then if any doubt I will ask u..and thank you so much sir for this much information😃