Dear Professor James Gaskin: I would like to express my gratitude to you for supporting researchers. Your videos, plugins, and estimands made our life easier. God bless you.
Thank you so much for sharing! This is definitely a lifesaver compared to having to conduct the indirect effects one at a time. I just found a small error that, when there are three or more paths, the standardized estimates would show incorrect values (also shown in your video) that are the same. It'd be wonderful if that can be fixed as I use mostly standardized estimates.
Hi James, on the last table. Line 2. I can see that the p-value is insignificant at 0.888 but the estimate is positive 0.121. That actually happened in my own analysis (p-value is insignificant and estimate is 0.389). My thought is if p-value is insignificant, estimate should be around 0.0xx. Does that really happen? I think I did not miss any step of doing mediation. Thank you so much James.
The estimate can be any value. The p-value is only significant if the amount of error in the relationship is low. Thinking of a scatter plot for a two-variable relationship, if the datapoints plot broadly, but generally sloping upward, then there will be a positive regression weight, possibly even of high magnitude, but the error will also be very high. The p-value is derived from the critical ratio, which is calculated as the slope divided by the standard error. So, even if the slope is high, if the standard error is high, the critical ratio will be low and the p-value will be high. Having said all of that, it is still more common to see smaller regression weights associated with non-significant p-values.
Dear Professor James Gaskin, first I want to thank you for always sharing videos with tones of useful information. Your plugins, estimand, and videos all have been a great resource for my research. One question I have is about assessing the local dependency of items in each construct when conducting the path analysis with latent factors. According to a study that I read, it says CFA Modification Indices could be used for detecting locally dependent items. I know that you usually demonstrates a path analysis in SEM with composite variables, but I wonder if you may have any idea with assessing local dependence with CFA - MIs when conducting SEM path analysis with the latent factors. I will appreciate it if you may give any comments or feedback. Thank you!
The modification indices in AMOS (and other cb-sem software) can identify dependencies between latent or observed variables when they are not already specified. In a CFA, this can identify which observed measures are highly correlated but missing parameters estimated between them (i.e., they're not explicitly correlated). I show that in this video: ruclips.net/video/sbTl-icEyD8/видео.html
Hello James, Thank you for your amazing videos. They help me a lot. I have a model with 6 independent variables, one mediator and 4 outcome variables. When I tested the direct effect of the independent variables on the mediator factor, they were significant. when I tested the direct effect of the mediator factor on the 4 outcome variables , they all were significant too. I did the mediation test to confirm the results by using your latent plugin. Surprisingly, I found that two of the independent variables do not have significant effect on the 4 outcome variables via the mediator factor ( their p value were more than .05 and less than .10) although their direct effects on the mediator ware significant (less than .05 p value) and the direct effects of the mediator on the outcome variables were significant (less than .05 p value) . What does this mean? and How can I explain that in my results section because the mediation test should support the results of direct effect relationships that have been confirmed?
Bootstrapping will often dampen the extremeness of p-values. So, it will bring them up from very low values. Aside from that, indirect effects are naturally small, due to multiplying two direct effects. So, we can tolerate smaller effects from mediation. Here is something I recently wrote about it (this has been accepted, but not yet published) "Effect sizes are influenced by the size of the regression coefficient. This is problematic because values of indirect effects are often small due to the method of calculation. Indirect effects are the product of two decimals. A decimal (less than 1) multiplied by another will always result in a number smaller than either original value. Thus, indirect effect coefficients are naturally small. Additionally, mediation models (particularly complex ones, such as parallel mediation) require both direct and indirect effects on the dependent variable and, therefore, more predicting effects (direct plus indirect) than nonmediated models (merely direct). This combined situation, in turn, deflates the f 2 (Lachowicz et al., 2018). Therefore, using Aiken’s or Cohen’s recommendations for effect size targets may inadvertently result in false negatives when applied to indirect effect sizes. Others have noted this problem but have not provided any practical alternative. For example, some (Ferguson, 2009; MacKinnon et al., 2007) have recommended that meaningful indirect (v) effect sizes may just require a larger sample size (e.g., > 500) to reduce errors associated with the mediated effect. Alternatively, Agler and De Boeck (2017), Lowry and Gaskin (2014), and Ogbeibu and Gaskin (2022) suggest that even if an effect size is considered too small, as is often the case with indirect effect sizes, it can still be useful to indicate the signal of an effect, and these signals can still serve as support for future predictive theorizing (Hofman et al., 2017; Liengaard et al., 2020). In an attempt to devise a solution to small v effect sizes, Lachowicz et al. (2018) suggest that researchers could square the specific v effects. However, this again reduces the size of the estimate (decimal * decimal). Ogbeibu et al. (2021) and Ogbeibu and Gaskin (2022) took this further by suggesting that the estimation of v effect sizes (v2) should be compared against a halving of Cohen’s (1992) effect size measure thresholds to compensate for inherently small v effect sizes. Thus, Cohen’s (1992) effect size thresholds of 0.02 (small), 0.15 (medium), and 0.35 (large) would be reduced to 0.01 (small), 0.075 (medium), and 0.175 (large) for evaluating the size of indirect effects. For further justification of these suggested thresholds, we point the reader to Ogbeibu et al. (2021) and Ogbeibu and Gaskin (2022). However, while suggested thresholds by Ogbeibu et al. (2021) and Ogbeibu and Gaskin (2022) make it more feasible to observe small and medium effects, it is still likely to be quite rare to observe a large effect. For example, to reach the large effect using their suggestion of 0.175, we would need the sample mean estimates for the indirect effect to equal 0.418 (0.418 x 0.418 = 0.175). But to achieve the sample mean of 0.418, the values of a and b (in Figure 1a) must be greater than 0.646 (0.646 x 0.646 = 0.418). Not only is this unlikely, observing coefficients a and b greater than 0.646 is probably an indication of multicollinearity. Thus, we suggest a more feasible guideline derived from a simple assumption that the coefficients a and b are small when 0.100, medium when 0.200, and large when 0.300 or greater. This translates into small v effect sizes of 0.01 (0.100 x 0.100), medium as 0.04 (0.200 x 0.200), and large as 0.09 (0.300 x 0.300). We next outline the approach to finding, calculating, and interpreting these effect sizes in SmartPLS." Here is the video about this: ruclips.net/video/pS5coOTcP-U/видео.html Here is the citation: Gaskin, J., Ogbeibu, S., and Lowry, P. B. (2023). Demystifying Prediction in Mediation Research and the Use of Specific Indirect Effects and Indirect Effect Sizes In H. Latan, J. F. Hair, & R. Noonan (Eds.), Partial least squares path modeling: Basic concepts, methodological issues, and applications (2nd ed.). Cham, Switzerland: Springer.
Hello Mr. Gaskin, when I use the Latent Mediation Plugin on AMOS 24, I only get an estimate value. No Lower, Upper, and P-Value. I also can't click Bootstrap Confidence to see those values. Can you help me? Thank you.
Many thanks Prof. Gaskin. Your estimand has been of great help! However, when I am trying to use it on a model with second order constructs; I am getting an error message " The given key was not present in the dictionary". Could that be because of presence of second order constructs or am I making some mistake?
Dear Professor Gaskin, thank you so much! I have a question, can we get standard error of the indirect effect by using the this method? Where could we find the results
Hello James, Thank you for your amazing videos. They help me a lot. Please, I have a problem when testing mediation. I have a latent model that contains a second order factor. I tried to use my indirect effect estimand (on Amos 24) and it did not work (Amos sends me an error message), although it works well with me when using it with previous model that contains first order factors only. Also, I need to know if Latent serial mediation estimand plugin work for Amos 24 or not? I tried to use it on the same model using Amos 24 and there was no values on the results schedule. Do you please tell me what is the best way for me to test mediation relationships in my case? Thank you in advance.
Using just an estimand should work just fine for a second order model. I'm not sure what might be causing the issue. The estimands also shouldn't depend on the AMOS version, however, the plugins only work for the versions of the folder they are in.
@@Gaskination thank you so much for your reply. Here is the error message that Amos sent to me when I use the estimands: An error occurred while attempting to fit the model. During the analysis of a bootstrap sample, an attempt was made to compute the correlation between two variables, one of whose estimated variances failed to be positive. This attempt was made because "Standardized estimates in the "Analysis Properties' window was checked, or because the Standardized method was used, or because a user-defined estimand requires calculating standardized estimates. Could you please explain what is this problem and how can I fix it to test the mediation relationships?
@@hebaelsayedelbdawyahmedhas8529 Sounds like there was a negative variance (which is impossible). If the output is available, check the notes for the model to see if it tells you which variance is negative. If you can identify it, then you can constrain it to be a small positive number (like 0.05) by double-clicking the variable or error/residual and going to its parameters tab and typing a small value in the box.
@@Gaskination Thank you James. I checked the notes for the model and did what you suggested. But, this message still exists " The following covariance matrix is not positive definite" and I can not run the mediation test. Please, how can I solve this problem?
@@hebaelsayedelbdawyahmedhas8529 It may be due to extreme non-normality of a variable (perhaps if you included a nominal variable directly, rather than through binary dummy variables). If you did include dummy variables, make sure to omit a reference category. If you have any variables that have zero (or near zero) variance (e.g., all male respondents and you included sex as a variable in your model), make sure to remove them first.
Hello sir, thank for your videos. i learned SEM by referring your videos only. I have one question. In SEM series videos you converted Measurement model into Sem by using cooks file. but in another video you make sem model by just arranging latent variables of Measuremnrt model, which one is more correct. is there are two ways to convert SEM model
Dear prof. James, Could I ask how to interpret the result on this video? I judged mediation occurs when the CIs for the indirect effects do not contain zero (Preacher & Hayes, 2008), but my results show that it does not contain zero but is not significant in p value, so should I judge the mediation effect using p value or CI? Thank you so much in advance.
Dr gaskin i have two questions when i use specific bias test both my unconstrained model and zero constrained model have same number of degree of freedom and same chi square value such that the difference comes =0 and p-value is 1. is it fine or am i doing some mistake. second question is : suppose our dependent variable is also a latent variable. do we have to include this dependent variable in EFA, CFA, and then impute the scores and finally get the imputed scores and form our causal model? can you please throw some light on it? regards
1. Version 26 of AMOS changed something in the code, and we are investigating this. It is our top priority at the moment. I'm hoping to hear back from my research assistant within the next few days. 2. The DV should definitely be included in the EFA/CFA if it is reflective and latent. Otherwise, you won't be able to determine the discriminant validity of the DV.
Hi How to connect dependent observed variable (rectangle) to independent latent variable (Oval)? If we connect it, will it not become part of the independent latent variable?
Mathematically, I think it should be the same. It still calculates the extent to which all other indicators of the latent IV are related to this observed DV.
@@Gaskination thank you very much Prof. Gaskin for your quick answer. Can I use these estimads for multiple group analysis as well? Thank you in advance
@@sabinaildkz3800 No, but you cold use the one called "MyGroupDifferences". You can find an explanation of it here: statwiki.kolobkreations.com/index.php?title=Plugins#LIST_OF_ESTIMANDS:
Hello sir, i have done the same way you have explained in this video. Thanks alot for this plugin sir. I have one doubt, the standardised estimates values are coming different. If we total unstandardized values, the answer is matching with the amos output but the same is not happening for standardised values. How will go by this as i am into reporting standardised values? Please help sir.
The plugin calculates the unstandardized indirect effect and then standardizes the result. So, the standardized result will not match the multiplication of two standardized paths multiplied together.
I've never used PROCESS. So, I'm not sure which estimand will be best. They are all described on the statwiki: statwiki.gaskination.com/index.php?title=Plugins#LIST_OF_ESTIMANDS:
Dr. Gaskin, is this plug-in by any chance available for Amos version 23? I have a serial multiple mediation in my research and I would love to try this plug-in. Thank you!
Unfortunately it is not. I have never tried it with version 23, so I'm not positive it won't work. You can try it. But I have no way to test it on my end. Best of luck to you.
Hai sir can I know what's the different between the result of using estimand "Latent Mediation" And "Specific Indirect Effect" ? Because I dont see any different for the result with the mediation of my Variabel with dimension/laten,, hehe thankyouu
Dear Professor Gaskin, I downloaded the plugin for Latent serial mediation from your Gaskination's StatWiki for amos Version 26. But the donloaded version is a .dll-file, which does not work for amos 26. I think it shoud be a .vb-file. Or am I wrong and I should keep looking to find the problem why the plugin does not run for me? Many thanks in advance.
Sometimes it needs one day. I found the problem and it is working now. By the way, your videos are awesome. They helped me pretty well through my PhD until now. Thanks a lot.
Unfortunately, this plugin does not allow that flexibility. I'll ask my RA to rebuild it with that in mind. However, my new RA is only one week into it, so he is still figure things out. It may be a bit. Thank you for the suggestion!
I don't know any place that it is legally available for free. Check with your institution to see if they have a site license. If not, then the least expensive place is probably the academicsuperstore.
Dear Professor James Gaskin: I would like to express my gratitude to you for supporting researchers. Your videos, plugins, and estimands made our life easier. God bless you.
You made my year 2020!, this will definitely simplify my subsequent studies . Thanks
Thank you so much for sharing! This is definitely a lifesaver compared to having to conduct the indirect effects one at a time.
I just found a small error that, when there are three or more paths, the standardized estimates would show incorrect values (also shown in your video) that are the same. It'd be wonderful if that can be fixed as I use mostly standardized estimates.
Thanks for catching this! I will let my research assistant know to look into this.
Hi James, on the last table. Line 2. I can see that the p-value is insignificant at 0.888 but the estimate is positive 0.121.
That actually happened in my own analysis (p-value is insignificant and estimate is 0.389). My thought is if p-value is insignificant, estimate should be around 0.0xx. Does that really happen? I think I did not miss any step of doing mediation.
Thank you so much James.
The estimate can be any value. The p-value is only significant if the amount of error in the relationship is low. Thinking of a scatter plot for a two-variable relationship, if the datapoints plot broadly, but generally sloping upward, then there will be a positive regression weight, possibly even of high magnitude, but the error will also be very high. The p-value is derived from the critical ratio, which is calculated as the slope divided by the standard error. So, even if the slope is high, if the standard error is high, the critical ratio will be low and the p-value will be high. Having said all of that, it is still more common to see smaller regression weights associated with non-significant p-values.
Dear Professor James Gaskin, first I want to thank you for always sharing videos with tones of useful information. Your plugins, estimand, and videos all have been a great resource for my research.
One question I have is about assessing the local dependency of items in each construct when conducting the path analysis with latent factors. According to a study that I read, it says CFA Modification Indices could be used for detecting locally dependent items. I know that you usually demonstrates a path analysis in SEM with composite variables, but I wonder if you may have any idea with assessing local dependence with CFA - MIs when conducting SEM path analysis with the latent factors.
I will appreciate it if you may give any comments or feedback. Thank you!
The modification indices in AMOS (and other cb-sem software) can identify dependencies between latent or observed variables when they are not already specified. In a CFA, this can identify which observed measures are highly correlated but missing parameters estimated between them (i.e., they're not explicitly correlated). I show that in this video: ruclips.net/video/sbTl-icEyD8/видео.html
Thank you so much. I am so glad I know you.
Thankyout. This is really gonna very helpful in future
Excellent
Wonderful!
Hello James,
Thank you for your amazing videos. They help me a lot.
I have a model with 6 independent variables, one mediator and 4 outcome variables. When I tested the direct effect of the independent variables on the mediator factor, they were significant. when I tested the direct effect of the mediator factor on the 4 outcome variables , they all were significant too. I did the mediation test to confirm the results by using your latent plugin. Surprisingly, I found that two of the independent variables do not have significant effect on the 4 outcome variables via the mediator factor ( their p value were more than .05 and less than .10) although their direct effects on the mediator ware significant (less than .05 p value) and the direct effects of the mediator on the outcome variables were significant (less than .05 p value) .
What does this mean?
and How can I explain that in my results section because the mediation test should support the results of direct effect relationships that have been confirmed?
Bootstrapping will often dampen the extremeness of p-values. So, it will bring them up from very low values. Aside from that, indirect effects are naturally small, due to multiplying two direct effects. So, we can tolerate smaller effects from mediation. Here is something I recently wrote about it (this has been accepted, but not yet published)
"Effect sizes are influenced by the size of the regression coefficient. This is problematic because values of indirect effects are often small due to the method of calculation. Indirect effects are the product of two decimals. A decimal (less than 1) multiplied by another will always result in a number smaller than either original value. Thus, indirect effect coefficients are naturally small. Additionally, mediation models (particularly complex ones, such as parallel mediation) require both direct and indirect effects on the dependent variable and, therefore, more predicting effects (direct plus indirect) than nonmediated models (merely direct). This combined situation, in turn, deflates the f 2 (Lachowicz et al., 2018). Therefore, using Aiken’s or Cohen’s recommendations for effect size targets may inadvertently result in false negatives when applied to indirect effect sizes.
Others have noted this problem but have not provided any practical alternative. For example, some (Ferguson, 2009; MacKinnon et al., 2007) have recommended that meaningful indirect (v) effect sizes may just require a larger sample size (e.g., > 500) to reduce errors associated with the mediated effect. Alternatively, Agler and De Boeck (2017), Lowry and Gaskin (2014), and Ogbeibu and Gaskin (2022) suggest that even if an effect size is considered too small, as is often the case with indirect effect sizes, it can still be useful to indicate the signal of an effect, and these signals can still serve as support for future predictive theorizing (Hofman et al., 2017; Liengaard et al., 2020).
In an attempt to devise a solution to small v effect sizes, Lachowicz et al. (2018) suggest that researchers could square the specific v effects. However, this again reduces the size of the estimate (decimal * decimal). Ogbeibu et al. (2021) and Ogbeibu and Gaskin (2022) took this further by suggesting that the estimation of v effect sizes (v2) should be compared against a halving of Cohen’s (1992) effect size measure thresholds to compensate for inherently small v effect sizes. Thus, Cohen’s (1992) effect size thresholds of 0.02 (small), 0.15 (medium), and 0.35 (large) would be reduced to 0.01 (small), 0.075 (medium), and 0.175 (large) for evaluating the size of indirect effects. For further justification of these suggested thresholds, we point the reader to Ogbeibu et al. (2021) and Ogbeibu and Gaskin (2022).
However, while suggested thresholds by Ogbeibu et al. (2021) and Ogbeibu and Gaskin (2022) make it more feasible to observe small and medium effects, it is still likely to be quite rare to observe a large effect. For example, to reach the large effect using their suggestion of 0.175, we would need the sample mean estimates for the indirect effect to equal 0.418 (0.418 x 0.418 = 0.175). But to achieve the sample mean of 0.418, the values of a and b (in Figure 1a) must be greater than 0.646 (0.646 x 0.646 = 0.418). Not only is this unlikely, observing coefficients a and b greater than 0.646 is probably an indication of multicollinearity. Thus, we suggest a more feasible guideline derived from a simple assumption that the coefficients a and b are small when 0.100, medium when 0.200, and large when 0.300 or greater. This translates into small v effect sizes of 0.01 (0.100 x 0.100), medium as 0.04 (0.200 x 0.200), and large as 0.09 (0.300 x 0.300). We next outline the approach to finding, calculating, and interpreting these effect sizes in SmartPLS."
Here is the video about this: ruclips.net/video/pS5coOTcP-U/видео.html
Here is the citation: Gaskin, J., Ogbeibu, S., and Lowry, P. B. (2023). Demystifying Prediction in Mediation Research and the Use of Specific Indirect Effects and Indirect Effect Sizes In H. Latan, J. F. Hair, & R. Noonan (Eds.), Partial least squares path modeling: Basic concepts, methodological issues, and applications (2nd ed.). Cham, Switzerland: Springer.
Hello Mr. Gaskin, when I use the Latent Mediation Plugin on AMOS 24, I only get an estimate value. No Lower, Upper, and P-Value. I also can't click Bootstrap Confidence to see those values. Can you help me?
Thank you.
Many thanks Prof. Gaskin. Your estimand has been of great help! However, when I am trying to use it on a model with second order constructs; I am getting an error message " The given key was not present in the dictionary". Could that be because of presence of second order constructs or am I making some mistake?
That's odd. That shouldn't happen with 2nd order factors. Try saving, closing, and reopening AMOS. If that doesn't work, I'm not sure what will...
Dear Professor Gaskin, thank you so much! I have a question, can we get standard error of the indirect effect by using the this method? Where could we find the results
It doesn't currently produce the standard error, but I can have my research assistant look into adding that.
Hello James,
Thank you for your amazing videos. They help me a lot.
Please, I have a problem when testing mediation. I have a latent model that contains a second order factor. I tried to use my indirect effect estimand (on Amos 24) and it did not work (Amos sends me an error message), although it works well with me when using it with previous model that contains first order factors only.
Also, I need to know if Latent serial mediation estimand plugin work for Amos 24 or not?
I tried to use it on the same model using Amos 24 and there was no values on the results schedule.
Do you please tell me what is the best way for me to test mediation relationships in my case?
Thank you in advance.
Using just an estimand should work just fine for a second order model. I'm not sure what might be causing the issue. The estimands also shouldn't depend on the AMOS version, however, the plugins only work for the versions of the folder they are in.
@@Gaskination thank you so much for your reply. Here is the error message that Amos sent to me when I use the estimands:
An error occurred while attempting to fit the model.
During the analysis of a bootstrap sample, an attempt was made to compute the correlation between two variables, one of whose estimated variances failed to be positive.
This attempt was made because "Standardized estimates in the "Analysis Properties' window was checked, or because the Standardized method was used, or because a user-defined estimand requires calculating standardized estimates.
Could you please explain what is this problem and how can I fix it to test the mediation relationships?
@@hebaelsayedelbdawyahmedhas8529 Sounds like there was a negative variance (which is impossible). If the output is available, check the notes for the model to see if it tells you which variance is negative. If you can identify it, then you can constrain it to be a small positive number (like 0.05) by double-clicking the variable or error/residual and going to its parameters tab and typing a small value in the box.
@@Gaskination Thank you James. I checked the notes for the model and did what you suggested. But, this message still exists " The following covariance matrix is not positive definite" and I can not run the mediation test. Please, how can I solve this problem?
@@hebaelsayedelbdawyahmedhas8529 It may be due to extreme non-normality of a variable (perhaps if you included a nominal variable directly, rather than through binary dummy variables). If you did include dummy variables, make sure to omit a reference category. If you have any variables that have zero (or near zero) variance (e.g., all male respondents and you included sex as a variable in your model), make sure to remove them first.
Hello sir, thank for your videos. i learned SEM by referring your videos only. I have one question. In SEM series videos you converted Measurement model into Sem by using cooks file. but in another video you make sem model by just arranging latent variables of Measuremnrt model, which one is more correct. is there are two ways to convert SEM model
I think this is the video you need: ruclips.net/video/dsOS9tQjxW8/видео.html
Dear prof. James, Could I ask how to interpret the result on this video? I judged mediation occurs when the CIs for the indirect effects do not contain zero (Preacher & Hayes, 2008), but my results show that it does not contain zero but is not significant in p value, so should I judge the mediation effect using p value or CI? Thank you so much in advance.
The p-valeu should be consistent with the confidence intervals. If not, then the CIs might be fore 90% instead of 95%.
@@Gaskination Appreciate your reply. I will have a look more!
Dr gaskin i have two questions
when i use specific bias test both my unconstrained model and zero constrained model have same number of degree of freedom and same chi square value such that the difference comes =0 and p-value is 1. is it fine or am i doing some mistake.
second question is : suppose our dependent variable is also a latent variable. do we have to include this dependent variable in EFA, CFA, and then impute the scores and finally get the imputed scores and form our causal model? can you please throw some light on it?
regards
1. Version 26 of AMOS changed something in the code, and we are investigating this. It is our top priority at the moment. I'm hoping to hear back from my research assistant within the next few days.
2. The DV should definitely be included in the EFA/CFA if it is reflective and latent. Otherwise, you won't be able to determine the discriminant validity of the DV.
Hi How to connect dependent observed variable (rectangle) to independent latent variable (Oval)? If we connect it, will it not become part of the independent latent variable?
Mathematically, I think it should be the same. It still calculates the extent to which all other indicators of the latent IV are related to this observed DV.
@@Gaskination Thank you. I was confused about this and was not able to find answer to this anywhere else.
Dear Prof.Gaskin, is there a way to get r square of each serial mediation or/and effect size for each effect? Thank you in advance.
Unfortunately, no, not unless you removed all irrelevant variables and ran separate models.
@@Gaskination thank you very much Prof. Gaskin for your quick answer. Can I use these estimads for multiple group analysis as well? Thank you in advance
@@sabinaildkz3800 No, but you cold use the one called "MyGroupDifferences". You can find an explanation of it here: statwiki.kolobkreations.com/index.php?title=Plugins#LIST_OF_ESTIMANDS:
Hello sir, i have done the same way you have explained in this video. Thanks alot for this plugin sir. I have one doubt, the standardised estimates values are coming different. If we total unstandardized values, the answer is matching with the amos output but the same is not happening for standardised values. How will go by this as i am into reporting standardised values? Please help sir.
The plugin calculates the unstandardized indirect effect and then standardizes the result. So, the standardized result will not match the multiplication of two standardized paths multiplied together.
@@Gaskination Thank you sir. It means we have to interpret results using unstandardized estimates only in serial mediation? Please clarify sir.
@@dr.anupreetkaurmokha5792 You can use either. Both values are correct. One is just standardized while the other is unstandardized.
@@Gaskination Thanks alot sir.
Dear Professor,
Kindly send me the AMOS estimands for Hays PROCESS Model 14. How do I run it I can't find any video relating to that?
I've never used PROCESS. So, I'm not sure which estimand will be best. They are all described on the statwiki: statwiki.gaskination.com/index.php?title=Plugins#LIST_OF_ESTIMANDS:
Dr. Gaskin, is this plug-in by any chance available for Amos version 23? I have a serial multiple mediation in my research and I would love to try this plug-in. Thank you!
Unfortunately it is not. I have never tried it with version 23, so I'm not positive it won't work. You can try it. But I have no way to test it on my end. Best of luck to you.
Hai sir can I know what's the different between the result of using estimand "Latent Mediation" And "Specific Indirect Effect" ? Because I dont see any different for the result with the mediation of my Variabel with dimension/laten,, hehe thankyouu
Correct. The results should be identical. Good.
Dear Professor Gaskin, I downloaded the plugin for Latent serial mediation from your Gaskination's StatWiki for amos Version 26. But the donloaded version is a .dll-file, which does not work for amos 26. I think it shoud be a .vb-file. Or am I wrong and I should keep looking to find the problem why the plugin does not run for me? Many thanks in advance.
Sometimes it needs one day. I found the problem and it is working now. By the way, your videos are awesome. They helped me pretty well through my PhD until now. Thanks a lot.
@@Songs-jh6ln Glad you got it to work.
Dear Professor Gaskin, if I use this plugin, could I set the bootstrap samples to 5000 at 95% CI and how?
Unfortunately, this plugin does not allow that flexibility. I'll ask my RA to rebuild it with that in mind. However, my new RA is only one week into it, so he is still figure things out. It may be a bit. Thank you for the suggestion!
@@Gaskination Thank you so much for your response. I've learned a lot from you. May you have a wonderful week!
James Gaskin, can you please let me know where I can find a free Amos as an add-on for my currently using SPSS v25?
I don't know any place that it is legally available for free. Check with your institution to see if they have a site license. If not, then the least expensive place is probably the academicsuperstore.
Hello sir,
I want to cite this estimand. Can you please provide me with the citation sir? Thank you
Gaskin, J. (2016), "Latent Mediation Estimand", Gaskination's Statistics. statwiki.gaskination.com
@@Gaskination Thank you sir
Respected Member kindly share the link from where i can download AMOS26 with crack
That would be illegal. So, I cannot condone it.