Structural equation modeling in free software JASP
HTML-код
- Опубликовано: 15 июл 2024
- In this video, I will demonstrate how to do structural equation modeling in free software JASP. Useful links:
JASP: jasp-stats.org/
Source 1: lavaan.ugent.be/tutorial/sem....
Source 2: www.routledge.com/Quantitativ...
Code 1
latent variables
ind60 =~ x1 + x2 + x3
dem60 =~ y1 + y2 + y3 + y4
dem65 =~ y5 + y6 + y7 + y8
regressions
dem60 ~ ind60
dem65 ~ ind60 + dem60
residual covariances
y1 ~~ y5
y2 ~~ y4 + y6
y3 ~~ y7
y4 ~~ y8
y6 ~~ y8
Code2
latent variables
ind60 =~ x1 + x2 + x3
dem60 =~ y1 + y2 + y3 + y4
dem65 =~ y5 + y6 + y7 + y8
regressions
dem60 ~ alpha*ind60
dem65 ~ direct*ind60 + beta*dem60
effects of interest
indirect := alpha * beta
total := indirect + direct
proportion := indirect/total
residual covariances
y1 ~~ y5
y2 ~~ y4 + y6
y3 ~~ y7
y4 ~~ y8
y6 ~~ y8
Much appreciated, Vahid jan! I can imagine how tiring it might be reviewing almost all major SEM concepts in under 40 minutes!
Really nice and clear lecture. Covered many statistical outcomes. Thank You for letting me understand how to use SEM in JASP
Thank you so much , very informative and easy to follow.🌹
Thanks for your teaching !
Thanks for this informative tutorial
Thanks very much, you r legand..... Really helped me a lot in my research
thanks for million time!! 🌻
Great video! Thanks. :)
Thank you😊
So now JASP opens Stata data sets (.dta). Thanks for showing that @ 3:14.
thanks! it very helpful❤
Hi! Thank you for this video, it really helps! I hope you do another one soon, with more variables ;)
Sure, I suggest you watch my Jamovi videos, too. Jamovi is also a free package.
Many thanks for this tutorial, really useful! One small question. If some of my variables are ordinal, while other variables are ordinal? I still shluld go for ULS or DWLS? thanks!
Thank you, this is very useful. My JASP version 0.17.3 does not give me the option to include the standardized parameter estimates on the SEM path diagram? The option is only available with CFA. Any suggestions?
Can you share syntax to calculate AVE and HTMT in JASP
Your tutorial video is extremely helpful. You explained it very clearly and simply. BTW, do you have any other tutorial videos or useful resources about how to modify the path diagram until the model is fit (p-value >0.05, NFI and RFI > 0.9, etc.)? Thanks in advance.
You can watch my recent videos on SEM in Jamovi.
@@VahidAryadoust
Thanks for the resources. Now I'm facing an issue again while running an analysis using the code 2. In this context, IM should be a mediator of TP, SP and CP to PL. However, after running the code, I got the error message. Could you please help look into this? Sorry for bothering again.
Thank you in advance.
# latent variables
TP =~ TP7 + TP8 + TP9 + TP10 + TP11 + TP12 + TP13 + TP14 + TP15 + TP16 + TP17 + TP18 + TP19
SP =~ SP20 + SP21 + SP22 + SP23 + SP24 + SP25 + SP26 + SP27 + SP28 + SP29
CP =~ CP30 + CP31 + CP32 + CP33 + CP34 + CP35 + CP36 + CP37 + CP38 + CP39 + CP40 + CP41 + CP42
IM =~ IM43 + IM44 + IM45 + IM46 + IM47 + IM48 + IM49 + IM50 + IM51
PL =~ PL52 + PL53 + PL54 + PL55 + PL56 + PL57 + PL58 + PL59 + PL60 + PL61
# regressions
PL ~ TP
PL ~ SP
PL ~ CP
IM ~ alpha*TP
IM ~ alpha*SP
IM ~ alpha*CP
PL ~ direct*TP + beta*IM
PL ~ direct*CP + beta*IM
PL ~ direct*SP + beta*IM
# effects of interest
indirect := alpha * beta
total := indirect + direct
proportion := indirect/total
Thanks a lot for this video Vahid ! I have a question : does it work the same with multigroup SEM ? I mean can we use the same code structure in JASP to have the good data ? Thanks by advance,
Augustin
When I write something on the lines of 'dem65 ~ ind60 + dem60', I get "lavaan ERROR: duplicate model element in...." How do I deal with it?
Hi Vahid, thanks so much for this wonderful tutorial. I'd like to know if JASP can calculate direct and indirect effects.
try the Lavaan's code for computing the effects. It should work. Otherwise, use Jamovi (see my recent videos, please).
Hello, This video is very useful. Thank you very much for it. I have a question for my research. The population is very large and there is no sampling frame. I'll use a survey questionnaire to collect data. I'll develop hypotheses based on the literature. So to test hypotheses I'll construct a model. But since the population is large, I have a time constraint and no sampling frame, I'll use purposive sampling along with snowballing techniques for the selection of samples from the target population. So can I use SEM to test my hypotheses? and what about PLS? What is the difference between SEM and PLS? What do you suggest for data analysis and testing hypotheses for my research? Any suggestion will be highly appreciated. Thanks.
In language assessment, CB-SEM is mostly used, but I suggest you use PLS-SEM for hypothesis testing. See Hair's works on PLS.
Thanks for you explanation. In my model, I have 5 measurement models, with three moderators in between the two main variables, so should I type like this indirect: alpha* beta*gamma*delta . Also, you mentioned that 0.578 is a pretty high variance in the total effect , is there any reference showing how much of variance should we consider high and how much consider low?
Hi for your first question, yes that should work out for you. Re the second question, please see the relevant chapter in the following book:
www.routledge.com/Quantitative-Data-Analysis-for-Language-Assessment-Volume-II-Advanced-Methods/Aryadoust-Raquel/p/book/9781138733145
Many many thanks for your explanation. I know it may be inappropriate to ask for, but may I know which chapter should I read for my SEM? Also, is it possible for me to see the statistical power of the models with JASP? or I should use another program, like SAS, to acquire the power of my model?
@@kaiyowu9044
Please see:
5. Structural equation modeling in language assessment (Xuelian Zhu, Michelle Raquel & Vahid Aryadoust)
I am not sure if JASP provides statistical power.
@@VahidAryadoust Many thanks, your videos are brilliant !!!
note 22:35
Is there any ulternative to sem for testing small sample size in likert scale (45 responsents only)
Perhaps PLS-SEM is a choice, but n=45 might still be too small.
hey, can you help me out to solve a problem
There is gender difference in a variable of my research. I wonder if I could put control variable in SEM? How can I do this in JASP?
how to GSEM
I wonder can I analyse moderated moderated-mediation in JASP? can you do a video? thank you.
That is on my to-do list.
@@VahidAryadoust please also do a non recursive model then😅😅😅
I'm not clear about the effect of interest. What does ":=" sign for?
That is a code to get JASP / R output the weights (e.g., alpha and beta and/or direct and indirect weights).
@@VahidAryadoust Thank you so much! The problem is when I use this code, the path plot shows some errors. Specifically, in this line, total := indirect + direct.
@@rahmanmonjur4928 did you check if you have both direct and indirect paths in your model? You might only have one of them (direct).
@@VahidAryadoust I don't see any indirect path. However, when I changed the operation (for ex, total := indirect * direct) then the path diagram appears again, although the indirect path is still missing.
@@rahmanmonjur4928 Since you do not have any indirect path, please do not include that line.