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!
Prof. Gaskin, These tutorials are really helpful, so thank you so much for them. I have however a concerning doubt on whether I should or not, use SmartPLS to analyse the data I collected. With a mixed factorial design, I have a between and within-subjects experiment. I've gathered that, with multigroup analysis I can analyse the differences between the 2 (between-subjects) groups, but how should I treat the rest of it(within subjects/repeated measures)? With 2 IV, 2 DV and 5 constructs mediating these relationships. What is your best advice on which software to use, or how to complement SmartPLS?
With experimental data, usually a paired sample t-test is used, rather than SEM. Although, we did use SEM for this expirimental study: papers.ssrn.com/sol3/papers.cfm?abstract_id=2534937
Dear Professor Gaskin, Thanks for uploading this informative introductory video. Please suggest, if we need to do scale homogenization before running smartPLS? Thanks
If you have multiple groups in your data, you can do a Levene's test before using PLS, or you can run an invariance test within SmartPLS: ruclips.net/video/MDgo2U_Qh6Q/видео.html
Thank you professor for replying. My query (considering your model presented in above video) is if I have collected data for skill acquisition on 5 point Likert scale, innovativeness on 7 point Likert scale and self-efficacy (CSE) on 3 point Likert scale. Do I need to transform different (5 point, 7 point and 3 point Likert scales) scales to a common scale before running the model in smartPLS?
No. They can be on different scales. That is fine. Three-point is small and may have some negative repercussions, but there is not much you can do about it now. You cannot transform it to have more intervals.
Hi James! Thank you very much for the good content. It helps a lot. However, I still have not figured out how to include control variables in my PLS regression. It would be great if you could answer that! Thank you in advance!
@@Gaskination Thank you for the helpful advice! However, when I coded the variable gender into two dummy variables - female and male, I get a Singular Matrix Problem when calculating. When I run the model with only one of the dummy variables, it runs. Do you have any suggestions how to solve it? Thank you in advance!
@@snezhinamilusheva703 gender (sex) is already a dummy variable because it is binary. Thus, if you code sex as 0=male and 1=female, then it is actually a female variable where 0="not female".
Hi Dr Gaskin, thank you so much for uploading these videos. I have general questions. One of the important conditions in CB-SEM is model identification, to be specific the model needs to be over identified for the fit indices to have meaning. Does the condition also apply to PLS-SEM? My second question is, in CB-SEM, it is a requirement to fix either factor or one of the item laodings to 1 for the software to be able to be able to provide estimates, is this a necessary condition for PLS_SEM?
The answer to your first question is no. PLS SEM does not require over-identification because it is not trying to minimize the difference between the observed and predicted covariance matrices. The answer to the second question is also no (or, at least, you don't have any control over parameter constraints).
Thank you so much Dr. Gaskin for your reply. Just a small additional question, what text book/article would you recommend for someone who is planning on using PLS SEM for them to read and for citation purposes?(Do you have yours?) I would also be very appreciative of you if you could make a video on assessing PLS SEM model fitness using all the fit indices required in PLS SEM. Once again thank you so much.
Here is a video for SmartPLS 3 model fit: ruclips.net/video/Lp3r5F-eVCs/видео.html As for a book, the Primer on PLS by Joe Hair et al 2017 I think is a good one. You are also welcome to cite my PLS article in IEEE: Paul Benjamin Lowry & James Gaskin (2014). “Partial Least Squares (PLS) Structural Equation Modeling (SEM) for Building and Testing Behavioral Causal Theory: When to Choose It and How to Use It,” IEEE TPC (57:2), pp. 123-146.
Dear Gaskin, Your tutorials are so very good! thanks for that! Yet I have a problem. In SPSS, I had to exclude some outliers from my data set and, to do do, I simply scored them as missing values so they would not be taken into account in the analysis. I then saved the data in the csv format and imported it onto Smart PLS3. However, in Smart PLS the data includes all the values in the analysis (also the ones I excluded as missing values because they were outliers). How can I transfer the missing values from SPSS to Smart PLS so that Smart PLS will not take them into account?
When you double-click the dataset in SmartPLS 3, there is a link called "Missing Value Marker" at the top of that tab. Click on the link for "none" and give it the missing value marker. For example, if it SPSS you designated -99 as the missing value, enter that same number here.
Neha Adnan you can set it up exactly as shown in this video, but without any mediating variable. There will just be one line between the independent and dependent variables.
Hi, SIr, I have created a mediation model in smartpls 3..but after calculation; report not showing Construct Reliability and Validity i.e. rho-A, AVE, etc
Those statistics are only relevant to reflective factors. So, if you have only formative factors, then they won't show up. However, if you do have reflective factors, then there is perhaps an error. I guess you could try PLSc to see if it shows up. Also, I don't think they show up after bootstrap, but only after regular algorithm.
On my survey, I have more than one questions on each indicators, how can I put these questions as one rectangle on the diagram? should I calculate the average of the responses first before I input the data?
I don't think I understand. Each factor (representing a construct) should be latently comprised of multiple indicators (usually corresponding to questions on a survey). If you want to collapse these multiple indicators into a single one, you can either average, sum, or create latent variable scores (see my video on 2nd and 3rd order factors for this).
@@Gaskination Sorry, what I mean is there are five tests which measure five variables the participants response on tests and have only one degree on the test (Total degree). So, every participant has five degrees. Can I use Path analysis using this Data via SmartPLS?
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!
Thank you, you literally save my research^^
Thanks for this video it really helped out
Hi .. thank you for all your videos… have you done on model selection/ comparison with BIC and GM ??
Here is all I have on model fit for PLS: ruclips.net/video/Lp3r5F-eVCs/видео.html
Prof. Gaskin,
These tutorials are really helpful, so thank you so much for them.
I have however a concerning doubt on whether I should or not, use SmartPLS to analyse the data I collected. With a mixed factorial design, I have a between and within-subjects experiment. I've gathered that, with multigroup analysis I can analyse the differences between the 2 (between-subjects) groups, but how should I treat the rest of it(within subjects/repeated measures)? With 2 IV, 2 DV and 5 constructs mediating these relationships. What is your best advice on which software to use, or how to complement SmartPLS?
With experimental data, usually a paired sample t-test is used, rather than SEM. Although, we did use SEM for this expirimental study: papers.ssrn.com/sol3/papers.cfm?abstract_id=2534937
Perfect! But I noticed that in that study you used AMOS, do you think it's 'smart' to try it with SmartPLS? Or too messy?
It can certainly be done with SmartPLS.
Dear Professor Gaskin,
Thanks for uploading this informative introductory video. Please suggest, if we need to do scale homogenization before running smartPLS?
Thanks
If you have multiple groups in your data, you can do a Levene's test before using PLS, or you can run an invariance test within SmartPLS: ruclips.net/video/MDgo2U_Qh6Q/видео.html
Thank you professor for replying. My query (considering your model presented in above video) is if I have collected data for skill acquisition on 5 point Likert scale, innovativeness on 7 point Likert scale and self-efficacy (CSE) on 3 point Likert scale. Do I need to transform different (5 point, 7 point and 3 point Likert scales) scales to a common scale before running the model in smartPLS?
No. They can be on different scales. That is fine. Three-point is small and may have some negative repercussions, but there is not much you can do about it now. You cannot transform it to have more intervals.
Sir @@Gaskination you are best.
thanks
Hi James! Thank you very much for the good content. It helps a lot. However, I still have not figured out how to include control variables in my PLS regression. It would be great if you could answer that! Thank you in advance!
Just include them like you would include any predictors. If categorical, use dummy variables as shown here: ruclips.net/video/L-uynRF8_cg/видео.html
@@Gaskination Thank you for the helpful advice! However, when I coded the variable gender into two dummy variables - female and male, I get a Singular Matrix Problem when calculating. When I run the model with only one of the dummy variables, it runs. Do you have any suggestions how to solve it?
Thank you in advance!
@@snezhinamilusheva703 gender (sex) is already a dummy variable because it is binary. Thus, if you code sex as 0=male and 1=female, then it is actually a female variable where 0="not female".
Hi Dr Gaskin, thank you so much for uploading these videos. I have general questions. One of the important conditions in CB-SEM is model identification, to be specific the model needs to be over identified for the fit indices to have meaning. Does the condition also apply to PLS-SEM? My second question is, in CB-SEM, it is a requirement to fix either factor or one of the item laodings to 1 for the software to be able to be able to provide estimates, is this a necessary condition for PLS_SEM?
The answer to your first question is no. PLS SEM does not require over-identification because it is not trying to minimize the difference between the observed and predicted covariance matrices. The answer to the second question is also no (or, at least, you don't have any control over parameter constraints).
Thank you so much Dr. Gaskin for your reply. Just a small additional question, what text book/article would you recommend for someone who is planning on using PLS SEM for them to read and for citation purposes?(Do you have yours?) I would also be very appreciative of you if you could make a video on assessing PLS SEM model fitness using all the fit indices required in PLS SEM. Once again thank you so much.
Here is a video for SmartPLS 3 model fit: ruclips.net/video/Lp3r5F-eVCs/видео.html
As for a book, the Primer on PLS by Joe Hair et al 2017 I think is a good one. You are also welcome to cite my PLS article in IEEE: Paul Benjamin Lowry & James Gaskin (2014). “Partial Least Squares (PLS) Structural Equation Modeling (SEM) for Building and Testing Behavioral Causal Theory: When to Choose It and How to Use It,” IEEE TPC (57:2), pp. 123-146.
Thank you so much.
Dear Gaskin,
Your tutorials are so very good! thanks for that! Yet I have a problem. In SPSS, I had to exclude some outliers from my data set and, to do do, I simply scored them as missing values so they would not be taken into account in the analysis. I then saved the data in the csv format and imported it onto Smart PLS3. However, in Smart PLS the data includes all the values in the analysis (also the ones I excluded as missing values because they were outliers). How can I transfer the missing values from SPSS to Smart PLS so that Smart PLS will not take them into account?
When you double-click the dataset in SmartPLS 3, there is a link called "Missing Value Marker" at the top of that tab. Click on the link for "none" and give it the missing value marker. For example, if it SPSS you designated -99 as the missing value, enter that same number here.
Sir, why does skillacq and CSE variables have some missing indicators like in skill acqusition its indicators are 1,2,3,4 and 7?
It is due to some cleaning of the dataset I did prior to this video in SPSS.
Hi. I need to know how to make model for independent varibale and dependent only. Ive one independent variable and its sub variables are four.
Neha Adnan you can set it up exactly as shown in this video, but without any mediating variable. There will just be one line between the independent and dependent variables.
Hi, SIr, I have created a mediation model in smartpls 3..but after calculation; report not showing Construct Reliability and Validity i.e. rho-A, AVE, etc
Those statistics are only relevant to reflective factors. So, if you have only formative factors, then they won't show up. However, if you do have reflective factors, then there is perhaps an error. I guess you could try PLSc to see if it shows up. Also, I don't think they show up after bootstrap, but only after regular algorithm.
On my survey, I have more than one questions on each indicators, how can I put these questions as one rectangle on the diagram? should I calculate the average of the responses first before I input the data?
I don't think I understand. Each factor (representing a construct) should be latently comprised of multiple indicators (usually corresponding to questions on a survey). If you want to collapse these multiple indicators into a single one, you can either average, sum, or create latent variable scores (see my video on 2nd and 3rd order factors for this).
James Gaskin i see. thank you very much prof.
Dear James Kindly tell how to do control variable test for demographics in Smart PLS
Here is a video for it: ruclips.net/video/L-uynRF8_cg/видео.html
Sir please describe which is your IV,DV and moderating veriable in your model
CSE is IV, Innovativeness is mediator, Skill Acq is DV. No moderator.
i am new to smart pls
after importing my data,
the indicator section shows "No indicators to show"
how do i go about it
This video might help: ruclips.net/video/qK05XYx5CwU/видео.html
Please , can i use only total degrees of tests in path analysis via SmartPls
I'm sorry, I"m not sure what that means.
@@Gaskination Sorry, what I mean is there are five tests which measure five variables the participants response on tests and have only one degree on the test (Total degree). So, every participant has five degrees. Can I use Path analysis using this Data via SmartPLS?
I"m very sorry, I still don't understand what you mean. What is meant by "degrees"?
anyone have Professional SmartPls sem licensed ?
SmartPLS does.