EFA Demonstration
HTML-код
- Опубликовано: 22 авг 2024
- This is a demonstration of how to do an exploratory factor analysis in PASW SPSS statistics. This demo was made for the DM students at CWRU, but may be useful to anyone seeking to learn to do an EFA.
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!
watching these ten years later! 100x more helpful than my lecturers, cheers mate!
I include all latent continuous variables (except control variables) in the EFA. You do not have to have a perfectly clean Pattern Matrix. My recommendation is to have loadings on each factor that average out to around 0.700 and make sure no cross loadings are within 0.200 (absolute value). So, if item1 loaded on factor1 at 0.730 and on factor2 at 0.500, that's fine. Hope this helps.
Clinical psych phd student here. THANK YOU for this video, and all your others for that matter. So helpful.
Principal Axis Factoring (PAF) is an extraction method which considers only common variance (places communality estimates on diagonal of correlation matrix). I've never heard of Principle Factor Analysis, but I have used Principle Components Analysis which is another extraction method that considers all of the available variance (common + unique) (places 1’s on diagonal of correlation matrix). PAF is preferred in SEM because it accounts for covariation, whereas PCA accounts for total variance.
That is a fabulous question, and the answer is: yes, but. Yes, you can certainly use the same data for EFA and CFA. But, if you have a sufficiently large dataset, you would achieve higher rigor by using two randomly selected subsamples of your data for each. This would demonstrate greater validity and reliability.
I recommend always doing an EFA, even with established items, because you are using them with a new context and new set of data. Doing them one at a time is problematic because when you move on to the CFA, you will still run into problems. However, if you simply cannot get the EFA to work, you may need to examine normality and outliers first. Often kurtosis will throw off the EFA.
Oh, haha. Initially, these videos were intended just for my students at Case Western Reserve University. So "next time" meant "the next time I see you in class", not, "in my next video". Sorry for the confusion.
U seem like a good and funny teacher, professor Gaskin thank u
I said to my husband thatI loveeeeee James Gaskin and he didn’t like that! Lol. Thanks GURU. Now I can’t wait to go back and run my stats.
will look at it now-- and thank you so much for all the help, it has been so difficult to sift through all the literature and figure out how to go about this
James, I love you. Best video ever
great video. helped me understand at least the core points of EFA. cheers!
It would be best to send it to your adviser. If you have a specific question, you can send it to me. But if you are looking for general advice, seek it from your adviser or those whom he/she recommends. Thanks! (I get about 30+ emails/requests per day, which I dutifully respond to...)
Hello! This may be beyond the scope of this video, but I initially did a Monte Carlo Parallel Analysis on my data in order to determine how many factors I should extract. After running an EFA (along with your example) and setting the fixed number of factors to 2, my 'total variance explained' chart says that both factors only make up approx. 34% of the variance within my data. You mentioned in the video that one should at least aim for 50%. What do you suggest I do with such a low percentage since my data should contain 2 factors?
I would look at the data from a theoretical perspective first, rather than statistical. Look at the variable content to see how many themes/constructs there are. Then constrain SPSS to extract that many factors. This should help. You can also look on the total variance explained table to see how many factors it would take to break 50% (even if they were not extracted during that run).
Thank you!
No. You use the EFA to inform the CFA, but you do not then need to then go back to the EFA. You are good.
I do not know of a way to do a 2nd order factor analysis in SPSS. If you want to jimmy-rig it, you could just do a first order analysis first, then create composites out of the first order factors (so that there is only one composite to represent each factor) and then do a new factor analysis with just the composites. This would be a quasi-2nd order factor analysis. I don't know of any literature that does this as a precedence. Usually 2nd order factor analysis occurs in the CFA.
I have a video to show how to do this. It's called "Imputing Composite Variables in AMOS"
Professor James,
Which source (either Book or Paper) do you use in order to stablish the threshold 0.2 to solve crossloading problems?
Gefen and Straub (2005) are even more liberal and suggest a difference of 0.100 is sufficient. Gefen, D., & Straub, D. (2005). A practical guide to factorial validity using PLS-Graph: Tutorial and annotated example. Communications of the Association for Information systems, 16(1), 5.
Thank you so much. I really appreciate it.
This is quite useful. I appreciate it!
The EFA is a very squishy process with no real systematic way of doing it. Usually if I have a set of data that is not loading very well, I will skip the EFA and will jump straight to CFA. I build the CFA based on what I expect. For example, I might expect there to be five latent constructs, and I probably know which items belong to each. Then I start trimming in the CFA based on the lowest loadings. This tends to simplify troublesome datasets.
Thanks Professor,
I really appreciate you. The god bless you...
Regads,
Mohammad
I don't have a reference for it. There probably is some reference out there somewhere, but I'm just speaking from experience.
Dear James
I have a question about EFA and second order CFA . At first EFA:
1) assume that we want to discover the relationship between two concepts with two separate questionnaire , for example customer satisfaction and employee engagement . Should we do EFA for each questionnaire (concept) separately or we should do an EFA for whole questions?
2)In second order CFA , we assume the example in your demo for second order structure, is it necessary to do a CFA for the model without second order factors with all first order factors? or we should do a CFA for the elements of second order factors separately and after that do a CFA for the model with second order factors?
Best regards
1. Do EFA with all first order reflective latent factors together.
2. Do the CFA with the 2nd order factor modeled. Some journals will require you to do a model comparison between the 2nd order model and a 1st order model, just to show that the 2nd order factor should be modeled as 2nd order.
Dear James
Thanks for your reply. for the last question, is it necessary to do CFA for every second order factor separately at first or its enough to do a CFA for the whole model?
You can do it for the whole model. I will often do a separate EFA, but not CFA.
Bravo!
It is very well explained, thank you!
Hi James.
Thanks for the videos, they are really interesting and clarify our doubts. I do not understand why it is better to use principal axis factoring in the factorial exploratory analysis and not the classical method of principal components.
I am currently doing this analysis as conducted under PLS my structural equations model and really I want to do a good analysis before proceeding. You could explain a little more about it, would greatly appreciate it.
Regards
They just serve different purposes. Principal Axis Factoring (PAF) is an extraction method which considers only common variance (places communality estimates on diagonal of correlation matrix). Principle Components Analysis is another extraction method that considers all of the available variance (common + unique) (places 1’s on diagonal of correlation matrix). PAF is preferred in SEM because it accounts for covariation, whereas PCA accounts for total variance. Nevertheless, an argument can be made for either.
Dear Dr. Gaskin
may I know that EFA is compulsory test for pilot study or we can do similar test with Smart-PLS, by checking the Chronbach's Alpha, composite reliability, outer loading, discriminant validity and convergent validity?
Thank you
You can also do it in SmartPLS. SmartPLS also produces a pattern matrix.
Thanks alot
If you mean multivariate outliers, then you need to look at the Mahalanobis d-squared. I have a video for this I think. Check my channel for an outliers video.
Hello James, Thank you very much for the videos. A quick question please. I have a 2 factor model and one of the items loads strongly and clearly (>0.5 and no cross loading) on the wrong factor. Theoretically, the item belongs to the other factor. Do I keep it on whatever factor it loads on and carry on? Or do I just delete it from the analysis?
Thanks
You can either delete it, or you can look at the wording of the question and how it compares to the wording of the other measures on that factor. If it is similar enough, you can argue that it makes sense why it is loading incorrectly, and then you'll still have to delete it I suppose. I would not keep it unless this is truly exploratory and you do not have prior literature establishing where the measure should go.
this is very helpful
thank you!
hi there i'm running an efa as follows:
Items with primary factor loadings ≥.40 (including values that rounded to .4) and secondary factor loadings ≤.30 and those that did not load on more than one factor were retained. Items not meeting these criteria were removed one at a time.
My question is - should i eliminate items in a particular order e.g. items with loadings lower than .3 should be eliminated before those with cross loadings higher? I hope this makes sense................
I think that this person means that he/she wishes to know which video lecture you recorded after this one, so that he/she can follow you in sequence. I too would be interested in this since you say that you will get into more of the details on EFA 'next time'...but looking through your video list, it is not really apparent to me which video corresponds to 'next time'.
Dear Dr. Gaskin
I have 5 high-order costructs in my conceptual framework. Do I have to conduct the EFA to all these constructs at the same time or can I conduct separate EFA to each latent variable separately?
I would conduct them separately for each higher order construct. Otherwise the lower order construct variables will all load on top of each other.
Your vidoes are fantastic. Thank you very much. Do you happen to have a version which would allow for testing multilevel effects? Thank you.
Great video. My question is 1) in my pattern matrix, items from different contruct variables load with Factor 1 (eg: q1, q8, q17 and p6 also load with Factor 1). However, actually the item questions focus on different aspect. Can I combine them into one factor or how do I adjust it? 2) some of the loading is negative, is it ok?
If items load together, but shouldn't, then you should see if there is some underlying concept between them. If not, then you might need to trim some of those items from the EFA. If a loading is negative, that means that it is correlating inversely with the other items on that factor. If this is unexpected, then it is probably because your item was reverse coded in the survey. You need to re-reverse the value so that it runs in the same direction as the other items.
James Gaskin thanks. Will try to adjust it.
James Gaskin I found that 1) there are nothing (no value there) of some items in the pattern matrix . What does that mean? 2) I have 10 construct factors but the SPSS suggests to have 15 factors. And the last two factors have loading with 2 items each. Do I have to delete them? Any suggestion?
Sabrina Chan I have a couple other videos on EFA that should help. The short answer is:
1. These items do not correlate with other items very well. Most likely you will need to drop them, but first try constraining the EFA to exactly 10 factors.
2. Try constraining first. Watch my other videos on EFA (part of the SEM Series and part of the BootCamp 2014) to learn more about dealing with problematic EFAs. At the end of BootCamp 2014 Day 3, I work through a particularly difficult EFA.
James Gaskin Thank you so much. I will try it out.
Dear James:
Thank you very much for aclarme doubt, you know some book that I can see to justify the use of the PAF method.
Nice to meet you in this way, the videos are really good and allow better interpretation of the results.
Regards
Chapter 3 of Hair et al 2010 "Multivariate Data Analysis" provides an in depth review of all these options in teh EFA.
Hi James, the determinant value for a data set is 1.412E-9. KMO, Barlet's test, and communalities are good. Is it appropriate to conduct EFA?
Yes.
@@Gaskination Thanks. I have one doubt, in addition. The factors loading range from 0.43 - 0.78. When I tried deleting items below 0.50 the determinant value and the scree plot look better (angle of inflection after the 3rd factor - initial 7 factors). KMO improved from .78 to .80. How would I justify deleting these items? If at all I should below 0.65, which I am not for. Thanks a lot. Your reply helped me. If you have a video on CFA on AMOS please share link.
@@anar2636 I have many videos on CFA in AMOS. Here are some search results from my channel: ruclips.net/user/Gaskinationsearch?query=CFA I would recommend the SEM Series 2016 as a good starting point. You can also check out the wiki page on this: statwiki.kolobkreations.com/index.php?title=Confirmatory_Factor_Analysis
As for justifying deletion of items... the best practice is to not delete items unless there is a lot of evidence that retaining them warps/distorts the measurement of the construct. When that is the case, you can state that you couldn't achieve minimum thresholds of validity without deleting them. Then you can cite one of the papers/books about validating measurement models: statwiki.kolobkreations.com/index.php?title=References#Measurement_Models
sir, i would like to use EFA to complete my research. My research is about tourism impact on economic, sosio-culture, and environment. on table of questionaire, should I separate economic variables(X1), socio-cultural variables(X2), and environmental variables(X3)? or don't separate it and let them find their own construct variables?
It is best to include all variables belonging to reflective latent factors in the same EFA. The only exception is if you have higher order factors with multiple subdimensions. Then you might consider running the EFA for just that higher order factor separate from other EFAs.
Hello James Sir
Sir, I have Little regarding Confusion PCA & Principle Axis Factoring Method
My Question is that if Developing new Questionnaire/ Scale of Any construct of Personality so which technique is used during the process of Factor Analysis.
Which Principle Component Analysis OR Principle Axis Factoring Method is used?
According to Hair et al 2010, you should try several extraction methods to see how they represent your data. The extraction and rotation methods are just a tool for better understanding the dimensionality of your data.
Thank you for the quick response-- another question I'm afraid.
ok so I use the EFA to inform the CFA and demonstrate good model fit ( and hence replicability). Is it acceptable to then compute factor scores ( for my entire data set ie the two split halves together) for furterh analysis, using my EFA model which I can do directly in SPSS? If not how do I go about computing factor scores based on the modified model dervied in the CFA using AMOS?
Dear Dr.James,
Thanks a lot for your very helpful demonstrations.
some researchers skip EFA for adapted questionnaire and they argue that EFA is just for developed measurements.
So could you please provide me with a reference for what you stated above because I want to cite that in my thesis. I am doing EFA for my adapted instrument because of the different context,subject and new set of data.
I'm not sure I understand what you mean. Do you mean you want to know which analysis you should do first, then second, then third, etc.? If that is the case, then go to my wiki: statwiki. kolobkreations. com The order on that site is roughly the order of analysis. Quickly summarized: data cleaning, EFA, CFA (including CMB), Mediation, Moderation, Interaction. Hope that helps.
Hi James....I have 127 items for 21 constructs ......Right now I have around 400 data points.........I am just confused how to do EFA on all 127 items ...as I have performed EFA on 200 data points but results are poor........How to go ahead?
Dear James,
I have 6 IVs with around 8 questions for each... When I do the EFA with all IVs together (48 items), lots of item will removed, but when I do separately (one variable each time), just two questions are removed... Is it wrong to do one variable each time (my justification to do so is, for case that questionnaire is adopt from others, no need EFA, but it is possible that one variable adopt from X, and another Variable from Y, so even for that case EFA is needed, which is skipped...
Hello James,
After the identification of factors through EFA,what should be done for using these factors for subsequent analysis like regression or SEM.Should we take the sum of the underlying items for each factor?
You could use the factor weights (sometimes called factor scores) in order to do some regressions. Factor weights aren't saved by default, but it is an option in the save menu during EFA. What I usually do is I do an EFA and then a CFA and then use the results from the CFA to do some SEM. I have a whole playlist on what to do in what order and how to do it. It is called the SEM Series.
Hello, I'm developing and validating a scale. Using split halves, I've first conducted EFA, and obtained what I think is a satisfactory model. I want to using the other split half conduct CFA with the aim of testing/ replicating the model that i obtained in the EFA. apart from achieving acceptable levels of model fit-- dos my CFA findings have to soemhow feed back into my EFA model ( ie improve it) before i go ahead and compute factor scores for further analysis-- hope this makes sense.
Dr Gaskin thank you for the materials you have made available. Pl am using a secondary data for EFA/SEM studies, does SEM run with secondary data. And some of the variables does not load on any factor, what do you advice me to do? thanks.
Khalu Josef Secondary data can definitely be used in SEM, but it must meet some initial assumptions. The measurement modeling (EFA, CFA) are only for reflective latent factors. If items do not belong to a reflective latent factor (i.e., measured by multiple interchangeable items) then they do not belong in the EFA/CFA.
What is the difference between "Principal factor analysis" and "Principal axis factoring"? When do I apply one or the other? I have this question for a long time now and I wonder what can I read to clarify this issues..
Hi James,
I found this EFA video to be very useful so I tried Dimension Reduction on my data, but when I ran it, it gave me these two errors, "Correlation Matrix : This matrix is not positive definite" and "Total Variance Explained: Extraction cannot be done. This extraction is skipped.
What am I doing wrong?
Thanks
Probably you have included a variable that does not vary. Sometimes this happens with categorical variables like gender or industry, and sometimes it happens by accident if you include non-sense variables like "completed survey". It might also happen if you have non-numeric variables in there, or if you have variables that are completely random.
Hi Dr Gaskin. I did EFA for 7 constructs and it looks fine (differences are more than 0.2). However, when I run all the constructs I have, it did not return any results. Is it because of too big data (143 items)? how should I solve this problem?
+Tran Phuong That is too many items. This is an indication that you have higher order constructs with multiple lower order dimensions. I recommend doing an EFA separately for each of these types of higher order factors.
Dear james
İ tried to do an EFA for a wellstablished questionaire in the literature before and some (exactly 3) of items went to factors except the predicted factor in the basic model and the loadings of them on the new factor is very high. what should İ do? should İ only eliminate that items from the model or İ should do something else? İ should mention that the items that are wrongly went, have no good conceptional relation with new factors.
hadi yasrebdoost That is unfortunate. This sometimes happens when the population sampled is very different from the original population, or when there are many bad responses from unengaged participants (for example, a respondent just puts 3,3,3,3,3,3,3... for every item). I recommend doing a careful data screening to remove unengaged responses, then look at kurtosis. Then redo the EFA. If there are still problems, then I recommend trimming away as few items as possible, but as many as are needed to produce a good result. Remove items only one at a time.
dear professor, i have a question.
i was wondering which correct EFA extracting method? 1) Put the All variable(Independent variable+Intermediate variable+dependent variable) or 2) Put the Independent variable once and intermediate +dependent once individually.
(for example in my thesis, 7 independent variables, 2 intermediate variables, 1 dependent variable)
and second question : why you don't using varimax? i wanna know difference between correct rotated method(varimax vs promax, equmax etc..). actually all the korean statistical book authors said, "just click varimax and they just recommend varimax method" they not consider other method. and then i wanna know seriously. thanks
best regards
professor..
+supremehh
1. All first order reflective latent factors should go in the same EFA. If some of the factors are 2nd order factors, they could go in their own EFA. If any of the factors are formative, they should not go in there. If any variables are observed (like age, education, etc.), they do not belong.
2. Varimax suppresses high loadings and inflates low loadings. Promax gives a "truer" more precise loading as you might see in AMOS. I sometimes use Varimax if I have extreme loadings.
+James Gaskin I appreciate for your reply. i'll use promax method in my thesis. really thanks so much.
thanks for this..
Dear Dr Gaskin
This is extremely a good video, very illustrative. I am asking if I can send to you my questionnaire designed for factor analysis for your comments. I am doing a PhD study on competitiveness here in Tanzania
Joh
Can you just give us an order list for these videos (a consecutive number for each video) please?
Thank you for your video. I would like to know if it is possible to consider the variance of the factors as a weighting. Thank you.
I don't think I understand what you mean. The EFA produces indicator loadings on factors. It also produces factor correlations.
James Gaskin
Thank you for the reply Mr Gaskin. I meant that I have 7 latent variables that I measure separately. Each variable has a set of 4 or 5 indicators (observed variables). Therefore I would like to know if the loadings can be utilized to generate a score of each latent variable.
Dear James,
Would you delete skew and kurtose items before conducting efa?
We've been taught skew = >|2| and kurtose >4 , not to use the standard error method.
Carin
Carin Marais yes, if you had enough other items after deleting. If items are few, then maybe be more hesitant to delete.
Thank you!
Thank you for the presentation, i have a questionnaire with 5 likert scales variables, 7 nominal variables. Can i only use the 5 likert scales variables to do the exploratory factor analysis? or choose 2 or 3 between them. Also, is exploratory factor analysis the same as convergent abd discriminant validity? I am a little bit confused. Thanks in advance for your reply.
You can have multiple sizes of scales in an EFA. Also, you can assess discriminant and convergent validity during the EFA. You can also do this during a CFA.
@@Gaskination Thanks for your reply. I am using SPSS software. Can i first doing EFA using dimension reduction command on spss, and then do discriminant and convergent validity for the validity test? Thanks
@@kossonouprunelle7576 Yes.
@@Gaskination Thanks a lot
Sir kindly help me out. While running efa, all factor loadings for one factor came out to be negative and have absolute value of more than 7. Subsequently, I transformed the values and reran efa and the model came out to be positive definate.
Please help. Shall be extremely for this.
When factor loadings are negative, it is usually because they are reverse-coded measures. For example, if asking about job satisfaction, you might include three measures (in this, the third one is reversed):
js1. I love my job
js2. I enjoy my work
js3. I dread going to work each day
In this case, the third one will load with a negative loading. To fix this, subtract that column's values from 1+scalesize. So, if using a 5-point Likert scale, subtact from 6. This will turn your 1 into 5, 2 into 4, etc.
@@Gaskination thank you sir.
Dear Professor,
I have 61 items.. How can I do the outliers test for all dimensions together? Boxplot check the outliers Item by item and can not be useful in this case.. Can I detect the outliers in AMOS and remove them, as it do simultaneously?
regards,
Mohammad
Hello and thanks a lot, I have 2 questions, please.
1- I have a question, what if the EFA giving me good results but it is mixing one of my independent factors with one mediator factor? It is putting all the questions related to those 2 factors in one column.
2- If the EFA showing a separate factor with only one question in its column, does it means I have to delete this question from the questionnaire?
Thanks a lot,
1. That means they are strongly correlated and are perhaps just two dimensions of the same thing. I would recommend doing a separate EFA with just those two factors and constraining it to extract 2 factors. Then delete the items with the highest cross-loadings. Then put the result back in with the larger EFA.
2. You can delete it or constrain the EFA to extract 1 less factor.
Thanks a lot,
Hello James, I ran EFA using PCA and varimax techniques. There are factors loadings in three column and some appearing in two column. How I can clean it? I am gonna do a CFA after this. Thank you so much.
Daisy G Here is a video where I work through an imperfect EFA: ruclips.net/video/X-O-OcJPCe8/видео.html
James Gaskin Thank you thank you so much, Your videos are all so helpful for my research.
Thank you very much! by the way, statwiki is awesome!
Hi Dr Gaskin,
I have 5 indicators and each of them is evaluated using 5 measures. I have entered these measures as variables in SPSS, how can I link these measures to corresponding indicators in SPSS? or How SPSS know each measure is linked to its proper indicator?
This can only be done in the EFA. If you want to then use these indicators in a regression, you will need to save factor scores during the EFA (see the save button in efa).
Thank you very much!!
Hello Prof.James, Can we do CFA instead of doing EFA?
I always do EFA, unless I have formative factors. EFA identifies discriminant validity issues better than CFA.
Hi, when I do EFA with a set of data, can I use the same data when I do CFA.
Hello Sir, if i survey by questionnaire which there are 2 likert scale with agreement and important under rating scale 1-5 as well so Could i run EFA in the 1 set ? Or i have to separate Qa and Qi analysis .Thank you .
I'm not sure if I understand your question. If you're asking whether you can run an EFA on multiple scales, the answer is yes.
@@Gaskination i am sorry yo make you confused sir so i have questionnaire with likert scale which one is agree and other is important(q1a-q10a and q1im-q10im) so Can i analyst by select all ? Thank you .
@@toeyvalsher6184 Oh, yes. That is fine.
If there are 2 constructs have items loading for both. what are reasons?
+Tran Phuong Usually this means they are too similar conceptually. They might be dimensions of the same thing. Try doing an EFA with just these two constructs' items.
Dear James. I want to do EFA and CFA for longitudinal data. The data is for 450 firms at three different years i.e 2007, 2010 and 2012. Can you please guide me about it.
+Syed Turab Haider Naqvi I haven't really worked with longitudinal data. If I had to take a guess, I would do the models separately for each year.
Thanks James. If we do models separately for each year than is it possible to compare them for discussion of results?
+Syed Turab Haider Naqvi Yes. The way to analyze the data really depends on your research questions and hypotheses. If you want to look at change over time, then you'll have to create new variables to represent this. If you just want to see if years are different in terms of certain variables, then you'll want to do an ANOVA. If you want to see if certain relationships between variables differ from year to year, then you'll want to do a multigroup comparison.
How can i judge that is it the significant value or not? In factor analysis
Factors do not have significance values unless you count the eigenvalue>1 as an indication of significance.
Dear Prof James.
First: Many thanks for your amazing videos. They help me much to learning on SEM.
For EFA, what if i all my observed variable are binary value (1/0=Yes/No), can I run EFA and perform its result to the next analysis like CFA on Amos ? Thanks Prof
+Lukman Manggo EFA is for reflective latent factors. Binary variables are rarely part of a reflective factor. So, I would recommend against EFA. Several binary variables that are part of a single construct are usually better aggregated as a score or index of some sort (like an average or sum).
Thanks Prof. I have more question please.
For example I have 50 binary variables that answered Yes/No as Independent Variable (X).
The Dependent Variables also have binary value Yes/No, the others as interval variable (Y).
I want to analyse using SEM technic.
I want to see the percentage of contribution on each X to Y, and the effect of each X to Y
What is your suggestion please ?
1. Can I run Multiple regression ?
2. Can SmartPLS handle like this model ?
Thanks Prof for your kind help.
+Lukman Manggo
1. When you have factors like this, they are not latent, so it is better to create a score or index.
2. After creating the score or index, you can include the resultant score as a single-item observed variable in either AMOS or SmartPLS.
+James Gaskin Thanks Prof. I am very appreciate it
Hi James.....while doing EFA ...in my analysis for few variables negative loading is coming....is it fine?
Try switching extraction or rotation methods. This will probably fix it.
How do you tell a test is unidimensional or multidimensional? Help please!
Sorry for the late reply. For some reason RUclips hid this comment from me and flagged it as spam. Multidimensional factors usually split apart naturally in an EFA that extracts factors based on eigenvalues. Another way to determine it is by examining the wording of the items to see if they fall under multiple themes.
Thansk a lot!
Hi sir,,,, i am trying to analyse my survey data in which my objective is to find the main factor due to which people do not prefer cycling…… in that case should i go for PAF or PCA?
mahim khan i’m not sure I understand your objective, but with EFA, the extraction methods are not terribly different. They are simply a tool to explore factors. So, I would recommend trying both and seeing what they tell you.
Ty so much for replying… actually i am new to SPSS and stats…….so my questions might sound stupid….I tried both of them… i am getting 7 component factors…. can i just limit my factors to 3 and disable Eigen value option…. will it be ok?
The difference between 7 and 3 is pretty big. It is an indication that you might have mulitdimensional factors. But to answer your question, yes, you can force it to extract exactly the number of factors you desire.
i watched your video "working through the messy matrix".... and finally my problem is solved....TYSM... can u make one video on "analyzing factors obtained from factor analysis"... i mean how do we know what these factors are telling us....
The factors are the groupings of similar items. The items that group together on a factor should share a similar theme or construct.
Hello sir please explain me what is meaning of rotation in here spss?what is purpose of it..what will be its effect on analysis
Rotation is the distancing (and tightening) approach for forming factors that are hopefully further apart from each other, but closer within. statwiki.kolobkreations.com/index.php?title=Exploratory_Factor_Analysis#Rotation_types