Structural equation modeling using AMOS
ฝัง
- เผยแพร่เมื่อ 29 ก.ย. 2024
- In this video, I demonstrate how to conduct a structural equation modeling (SEM) analysis in AMOS. As SEM is based on confirmatory factor analysis (CFA), I would suggest you watch the following videos:
Video 1: • Confirmatory factor an...
Video 2: • Confirmatory factor an...
You might also want to read the book chapter mentioned in the video:
www.routledge....
Thankyou very much for sharing this and all your knowledge on youtube. Youve helped lots of people. god bless
This was quite useful. Just so we can replicate your analysis, can you please provide us with your sample data spss file? Thank you.
The data are the property of my university. I am afraid they cannot be shared.
Very very effective videos to learn AMOS easily. Thank you sir.
My cmif value is 6.96 .what can I do
Hi Dr, This is Alex, a research assistant from Malaysia. I have a question about this SEM. Before that, thank you for your excellent and helpful video! I am currently looking at a construct of a model and there are a few predictors that are dichotomous and dummy variables, so my question will be can SEM incorporate dichotomous and dummy variables in the estimation? Thank you, Dr.
Yes, you can. Please watch the video on MIMIC SEM analysis. I have explained how these variables can be used in SEM analysis.
th-cam.com/video/V473CcoVAvw/w-d-xo.html
@@VahidAryadoust Thank you so much, Dr. I have another question. I just estimated my model and my Chi-square is significant. Theoretically, it should not be significant. However, my GFI, AGFI, CFI, and RMSEA showed that my model is fit. So, does this imply my model is still fit on the data given the Chi-square is significant? Thank you Dr.
@@alextan806 the Chi-squared index in SEM is sensitive to sample size. I would rely on CFI, and TLI (NNFI) and RMSEA, plus normed chi-squared (X^2/df < 3).
@@VahidAryadoust Yes, there are some arguments saying that Chi-squared is sensitive to sample size where a large sample size will lead to significance in the test, I have 5000+ observations and I think this explains why the Chi-squared becomes significant. Thank you, Dr. Your video is so helpful and I really learned a lot from your video~
Thank you very much my dear brother, your way of lecturing is so interesting and attracting. ...(Can I access for AMOS app for free, please share the info).
Sir, in SEM analysis when I calculate the estimates it shows me that "The variable, M1, is represented by a rectangle in the path diagram, but it is not an observed variable". please help... i am not able to proceed further.
Thank you very much for the valuable insights towards SEM creation with IBM AMOS. It was really helpful!
Could u please provide the dataset?
14:01 hello , when I try to run the analysis, pops out “the variable , YXQ3 , is represented by a rectangle in the path diagram, but it is not an observed variable”. What can I do about that? Thank you for the answer in advance ❤ have a nice day 😊
Hello Dear, wanted your advise regarding the following issue I am facing.
When there is no covariance amongst the the independent / exogenous constructs (when i don’t draw it), the model fit is poor. But when the covariance line is drawn among the independent construct the model fit becomes excellent.
In your video also the model 2 depicted a better result than model 1. What are the difference between these model? Based on which model's result should state my model fitness.
I suggest that exogenous variables should be correlated / covaries. This will have several advantages, including better model fit.
@@VahidAryadoust hello dear. Thanks for replying.
But isn’t it a problem that without covariance line the model fit is so poor but with the lines the fit shows excellent result?
.
One more problem that I face when I draw the covariance line is after drawing the line the measurement model fit indices and structural model fit indices shows the same value. Shouldn’t those values be different?
@@salmapapia9737 IF the model without covariances is a poor fit, then it is a wrong model. but you should still report this in your paper. The next for you to take will be to include the covariances in the model and estimate the fit and parameters of the models. I assume the model's fit will improve after adding the covariances. If this does not help, I can take a quick look at your data. Please email it to me/get in touch via my linkedin please.
Thank you, for the interesting Lectures, however, I faced one problem when I calculated the model the error terms were not displayed in the path diagram? therefore I kindly request your help. Regards
Thank you Vahid, that's very helpful. In terms of calculating the number of free parameters, I noticed different sources use different formulae in estimating a total number of structural model parameters. Any suggestion?
Sir is it possible for you to do sem analysis for research scholars?
Hello. Should Likert Scale in the questionnaire not be treated as Ordinal data?
Respect Sir, This is so helpful. Thank you to make this video
This is nice one Dr. Vahid Aryadoust.. I really appreciate the steps you demonstrated in this video.. Can I mail you and request for more assistant please?
sure, if I can be of help.
Thanks Dr. I was able to get best model....Just would like to ask if need to have the reference after Amos suggest to covariate....
How can I reach you thru email?
Thank you so much
I am getting an "NaN" Not a Number error in Master Validity table. I checked for missing data, assigned values but it is all accurate. Can you please provide any guidance?
Thank you for sharing this. I just have one question, what if in my hypothesis there are variables that negatively effect another variable. Is it still the same as in the video?
If that is theoretically justifiable, that is fine. Otherwise, they might need to be reverse-coded before being entered into the analysis, or removed, etc.
HELLO SIR I HAVE APPLIED THE SAME STEPS BUT AFTER PRESSING THE ANALYSIS OPTION I GOT MESSAGE THAT C1 IS PRESENTED INTO BOX BUT NOT OBSERVED VARIABLE. WHAT SHOULD I DO ???? PLZ EXPLAIN
Sir, may I have your email address please?
Thank you Sir.Sir can you please tell me when we obliged to use SEM than other models ?
Thanks for your excellent explanation. Where we can get the data set used in this presentation?
Can I have this sample data
Dr, after watching your video and i conducted a SEM analysis and my supervisors asked why "two standardized estimates more than the unit" I really do not understand what he meant.
Hello, do you know why the rotate or reflect commands don't work? I am using AMOS 27.
Sir i am working on Amos from last 2 month on the same model but my p value is not significant toh is case main data ko kaise set kr skte h
I have a question that you didn't use the Composite Factor Scores for Latent Variables to draw the Structural Model.. but you built the model along with Observeable items of Latent variables.
Mostly authors persists on using the Composite Structural Model by imputing Factor Scores of Latent items. What's your point of view?
Hi Dr. Can yu make a video on how to prepare datasheet for Amos. It will be very useful. How to enter data in dataview. Thanks in advance
Dr. good evening, what do I do if the standardized values turn out to be very large values like 12.25 or 35.45?
Sir is this CB-SEM?
Hi sir. When you switch to model 1, which was already prepared by you, there are variables. How and why did they appear there?
Well explained, better than some other videos. Possible to post the data file? Without that it is not possible to try in Amos (and jamovi to cross-check) and perhaps dilutes the usefulness of this video.
thank you so much for the video, can anyone explain me if I can use likert scale data for SEM as independent variables? I saw it in the video that the data is also in likert scale. is it possible in SEM ?
When I run the analysis, I am unable to find the values nor view the output. Please suggest sir.
Thank you for this video which really helps a lot. I am wondering that if we draw covariance based on the suggestion from Modification Indices, what are the meanings to do so? This absolutely improves the model fit, but the model is the same and the items are the same. If we report the statistics, what are the differences between the original model and the updated model drawing covariances? I appreciate your help if you could answer these questions. Thank you.
Covariances could indicate local dependence (bad news for the model) if they are between errors of items on different factors. On the other hand, covariances between factors indicate an association between them.
@@VahidAryadoust Thank you so much. This is very helpful.
Hi Dr, why my regression value is more than 1.0? Is it correct?
You might be looking at the non-standardized regression weight, which can be rater than one.
@@VahidAryadoust Thank you..Lete check again...Any ways to reach you aside from here?
Wonderful
Thank you for this video. Could you please guide me on how to test the moderating effect of variables on the relationship between the dependent variable and the independent variables? Also, I would like to understand how to find the effect of the Control Variable on the dependent variable.
Thank you, Dr.
I will discuss that in a separate video.
Hello Sir and thanks for the video, in the regression weights i can see only estimates, it does not show up SE CR and P values , any help? Thank you.
sir what is the value marked on top of the rectangle is that varience
Hi, love the video. I just wanted to ask whhen the estimate is 1 and there is no p value, what does it mean?
it is natural. when beta is fixed to 1, p values are not estimated.
@@VahidAryadoust Thank you for the response! does this mean that it is significant or not?
@@marthapedroza174 To find out, fix another path to 1, and remove '1' beside this path; re-run the analysis to see the p value.
Is it allowed to use observed variables in structural equation models directly without naming a latent construct? I found a paper citing hair et al. 2009 for the same. Couldn't find any video or article stating support for the same.
Yes, it is fine.
Very informative video, thanks for rhe efforts. I have a question, could i use SEM for ecological studies. For examples, to relate different factors with species diveraity data.
Generally, if your data meets the requirements of the models, why not?
If we've two mediators and both of their (mediating and moderated mediating) effects are insignificant, then is it considered to be a wrong model? Do we MUST have a mediating and/or moderated mediating effect in our model?
The model indicates no significant mediating/moderating effects. Look at the fit stats to decide if the model is suitable for the data.
Hi Dr. Vahid Aryadoust, Thanks a lot for your channel and TH-cam Videos. Please allow me to ask a stupid question.
I have 6 IV’s and 1 DV. I ran CFA and was able to get good model fit. Now struggling with Path Analysis.
As I run independent (IV) to depended (DV) arrows, AMOS tells me model is un-identifiable.
However, when I run it as DV to IV errors with error terms as bellow it works.
Looking for some guidance and direction to successfully use SEM
Regards
Ahsan
"the model is un-identifiable" means there are too many parameters to estimate. Remove some of the parameters, e.g., correlation etc to be able to run it again.
Thank you very much. Can you provide the data set please for practicing or referring us to useful and downloadable data sets?
The data are the property of my university. I am afraid they cannot be shared.
If the estimate is less than 0.5 and p-value is more than 0.05, should we remove the variable from the model and re-run the model? Will it improve the results?
Yes, you should.
hello sir i hope you are fine. sir i am working on pathways leading to certain outcomes. Like, one factor leads to some other factors and those factors also form another tree of factor. Lastly leading towards some outcome construct. Is this possible with SEM? It is more likely a propagation of factors. Please help
It could be solved with an SEM model, depending on the sample size and variables etc.
your model is not fit as it not shown GFI AGFI
Respect sir, can you help me please how to represent the mediation and direct effect in graph?
Hi doctor sorry for bothering you up with my questions, i have 8 latent variables 4 independent and 4 dependent,
What i understand from this vid and the two others for cfa is that i should first build and test the relationship for for each two independent variables ?!
and if they do not fit i shall not include the in the SEM ?!
Is that what you mean doctor ?
First, build one CFA per each latent variable. So, in your case, build 8 independent CFA and run the analyses separately. Once you are sure they work fine, include them all in the final SEM analysis.
That was very helpful...thank u so.much 😊
Thank you for these informative videos. I have a question concerning AMOS, when I draw the variables there is always 0; just above the circles, is that okay or not, because they did not show up in your explanation, just 1 above the arrows. I use version 26. Thank you!
It is most likely the mean of the variable set to zero. If you have chosen 'estimate intercept and mean' (due to missing data), you will see that. It is okay
Is there a way to highlight the significant paths in the visual model (e.g., with *, **, ***)?
Enter them manually.
Hey Dr.
Thank you for your time in explaining this. Just a quick question. I have checked the other videos on CFA. You have done the same analysis in both CFA and SEM videos. I am a bit confused on what is the difference between these two? I have been asked to complete these two analyses as part of my Thesis, the point is, in both videos you have done the same analysis? thank you so much for your time.
CFA does not include regression paths between latent variables, whereas SEM includes such relations.
Why is there a 1 on one of the paths from the construct to the indicator and how do I choose which regression weight will be 1? Thank you!
It is necessary to fix one of the paths to 1 in SEM analysis. This is done for parameter estimation. In Amos, it is automatically fixed to 1, provided that you follow the steps in the video.
Very helpful, thank you very much!
But is it also possible to select cases / set filters? I have to calculate SEM's for women and men! And I let SPSS run in the backround to do so, but AMOS took nevertheless all cases. 😞
Pls see my videos on multigroup / multilevel SEM.
@@VahidAryadoust Thanks a lot!
Thank you for the video. Do we need to do the validation if we use the pre-validated measures developed and validated by other researchers?
Yes, it would be necessary.
How does one interpret error terms and intercepts? Thanks.
Thank you, it was very useful!!
Thank you so much! Great explanation. Could you share the link here on the two videos you are specifically suggesting at the end of the video? You have quite many so I could not yet figure out which one you meant. Thanks a lot!! Best wishes
sure, here is the list of the videos:
th-cam.com/video/HKs9vIkpIXE/w-d-xo.html
Thank you so much.Will you be able to do a separate video on analysis of moderating and mediating effect
sure, I will note it down on my to-do list.
Doctor,
Your help please , i have a problem , some constructs are having AVE less than 0.5, i thought it is bcz i insert zero for the " not aaplicable- NA" in likert scale 🥺
Your opinion please 🙏🙏🙏🙏🙏
NA could be the cause. You might want to remove the NA cases, because they do not measure anything for you.
@@VahidAryadoust thanks doctor 🙏🙏🙏
Hello, Vahid! Are you providing private tutoring online ?
I am afraid I don't.
Hi Dr. Vahid. Is there a way that I can send you an email with regard to this presentation? I am currently working my dissertation paper and I believe that your expertise will aid me in completing it. Let me know how I will be able to reach you. Thanks a bunch!
Thank you for the video! it helps me so much, I have a question about the factor loading, is there any way to calculate it manually? if that so, how can I do it? thank you
YEs, there is, but it will take for ever.
@@VahidAryadoust I see, is it just the same as analytic factor? the concept? or it’s different?
Amazing Video! Thank you.
Hi Dr Vahid, How do I test within AMOS a 6, 5 , 4 , 3 factor solutions as a nested
version of the 7 factor CFA solution with all of the items used to generate the
7 factor solution.
Hello, Dr. Thank you for your helpful videos.it helps a lot in my research. I have a problem can you help me out. Whenever I go to calculate the estimates, it shows me an error. (No list of input variable was found. A list of input variables must appear once for each group). I already check the spss file and load the file into AMOS. but still, I got the same problem. what should I do now?
I suspect the problem is that the data is not read by AMOS. Please ensure the name of the variables in the model and those in the variables list are exactly the same.
Sir, Is it necessary to have SPSS Statistics installed to do analysis in AMOS?
No, it is not necessary to have SPSS.
I also want to study the customer satisfaction towards andriod phone
I have 5 independent variables
1)Product
2) quality
3) price
4) promotion staragies
5) feature
And one dependent variable is customer satisfaction (Y)..
Confirmatory factor analysis is done
But I don't understand how can apply the structure equation modeling then???
Can you please help me???
Your independent variables should have arrows moving to your Y variable, as explained in the video.
@@VahidAryadoust
You mean
I find correlation same as you find in this vedio??
Thanks alooootttttt
You are the best and also your method of teaching is also superd
@@faizanaqvi4437 Yes, I suggest you follow the steps in the video. I will soon release another video on SEM.
@@VahidAryadoust
DNT understand
What I do?
😔
I also want to study the impact of planning on success which is my main hypothesis
Could I get some sample files from spss to practice?
Check this website out: www.iris-database.org/iris/app/home/about
I have watched it before and i am watching it again. thanks a lot
Thank you so much Sir 👍👍
Thank you so much for this video.
sir, Certainly the impact of your help is indisputable, we learn a lot from you. just wanna ask in the SEM model, is it necessary to plot the covariance between the independent variables?, cuz when i did, results show no significantion to dependent variable. is bcuz the data?.
Drawing covariance arrows should be informed by theory. It can also help improve the fit of the model in most circumstances.
@@VahidAryadoust Totally OK, I have no problem with model fit. but how can I fix signification problems, I didn't find that any independent variable has significant relationship to the dependent one?.
I think my survey was not clear enough or bcuz reckless in answering the question. am totally confused. thanks in advance
@@zakihoussam2507 1. Remove the paths that are not significant 2. Remove the items that are not significant 3. Alternatively look at the revisions suggested by AMOS to revise the models (this will be provided only if you do not have missing data).
@@VahidAryadoust I truly appreciate your help and time you spent to answer questions.
definitely will follow your instructions 🙏👌
You video is amazing
Sir, do we need to put covariance in our independent variables in structural model even though in our hypothesis/conceptual model we do not correlate those?
I am very new to SEM... The question I had in my mind is that in structural modelling, do we include or delete the items that we had previously deleted during the measurement model validation? Or just like start from the scratch by listing out all the items as we had done before while constructing a measurement model? Please help! Also, a big thanks for the insightful video
Delete the items that did not function properly during the measurement model stage.
@@VahidAryadoust Did that and used that model as the structural model but could not achieve construct validity... Do I just interpret the model fit indices?
@@tantantan1771 Construct validity is a complex issue. A SEM analysis is not sufficient for establishing it. See:
journals.sagepub.com/doi/full/10.1177/02655322211018139
Meanwhile, you have not been able to find a reasonable model-to-data fit; if the model cannot be improved anymore, typically, the fit etc are to be reported.
@@VahidAryadoust Fit estimates seem to be fine but i am not able to interpret the data... My latent variables aren't explaining any effects over each other... There's no significant effect of the selected constructs over the dependent one whatsoever (got this info in the estimates):(
Is it possible to have the same model fit for both measurement model and structural model?
Yes, the TLI and CFI etc. can be the same for both.
@@VahidAryadoust thank you for this quick reply 🥺.
I have a paper submission deadline tomorrow and I am literally struggling. Another last question, do we need to give a reason for co-varying factors in the structural model because I haven't seen many models where factors are covaried. I have three constructs of impulsivity and model fits better if I covary them. Any citation reference if possible?
Hello sir, how can we access the dataset of this lecture?
The data cannot be shared. Sorry.
Excellent work!!! Congratulations!!!
Thanks for the video sir.
I noticed you said one would need to confirm the 4 constructs in separate models while it’s my understanding that it’s the overall model fit that is a requirement for SEM. In most papers I’ve come across the author has calculated the model fit indices for the entire model rather than reporting it for each latent variable separately. The composite reliability and convergent and discriminant validity are of course calculated separately for each factor. I may have misinterpreted what you’ve said so any comment or clarification is much appreciated. Best regards.
The composite reliability and convergent and discriminant validity should be reported, but also the fit of the measurement models, too. Examining the fit of measurement models is not commonly done, but it would be pretty important to carry out that step.
Many thanks Professor 🙏
Good video on sem Dr. Can I ask you one doubt..that can I do structural equation model using variables with different likert scales..
Generally, yes, you can.
@@VahidAryadoust If you can please make a video of it. It will be very usefull to social science students. Just a humble request.
@@rohts1 I will soon release a new video on SEM. Please stay tuned!
@@VahidAryadoust thank you
Non-sense, how one measure stress in Number???
Hello i have this error in my statistiques analyse can you help me please
tqvm sir
how can download Amos for free? pls
Yes you can. Check out the IBM SPSS website.
Hello, excellent explanation of SEM. Do you think you can share the data?
Sorry, the data is the property of the National Institute of Education, NTU.
I have two constructs.One is planning which is independent construct/variable and other is Success which is a dependent construct consists of two items(successful completion of project in terms of cost and successful completion in terms of schedule). CFA has been carried out and model fit is attained.I want to study the impact of planning on success and also impact of planning on schedule and cost separately.Can I do this by regressing schedule and cost on planning/independent construct separately after regressing success/dependent variable on planning?
Yes, that sounds fine.
Thank you! Your help is highly appreciated.