Hello Dr. Gaskin, Thank you for your video. What citation do you use to support the cut offs you use for goodness of fit estimates? (e.g. goof it is RMSEA 0.90, etc.) Thank you!
Dr. Gaskin, thanks for the video, I am following the steps you have shown in that video for a measure that has four subscales. My output does not produce fit stats (RMSEA, chi square etc), can you advise on why it might be happening?
This is a very helpful video James. Have you figured out is there a simple way to extract CR, AVE and construct correlations from Mplus? I guess no excel stats package is available yet like with AMOS? Thanks for your help and keep up the good work!
Thank you, it was very helpful! One question, do you know the reference you used to say that the correlations between factors should be above .8 or at least above .9? Can't find the .9 criteria!
Sorry about that imprecise language. I've learned more since making this video. Here is an updated one with more standard measures: ua-cam.com/video/4LGQ_t0xuj8/v-deo.html
The video was really helpful. I am just confused about STYX and STDY (which one is used for categorical var, as I got y for categorical and same for YX). plz clearify this.
Hello SIr, I have used AMOS 21 where Confirmatory factor analysis was calculated where i got Chi-square = 528.338. Tell me what is the acceptable value of Chi-square.
chi-square is relative to the number of variables and the sample size. So, a complex model with a sample size of 500 may have a chi-square in the 10,000s, but a simple model with a sample size of 75 may have a chi-square in the 100s. The general rule is that you want to minimize chi-square. A good relative measure is the chi-square/degrees of freedom. This should range between 1 and 5. Better measures are the SRMR, CFI, and RMSEA though.
It should work through a virtual machine (like citrix). I'm not sure if it works directly on the new Mac. You'll need to reach out to the Mplus creators to check.
Hello Sir, whenever i perform the CFA, and calculate the estimates, the window show that "The variable is represented by a rectangle in the path diagram, but it is not an observed variable". tell me why it is happened again again and i have observed that my data is correct.
This means that the dataset you are linking does not include a variable with the exact same (case and spelling) as the variable you are trying to model. So either the dataset is incorrect, or the syntax includes a spelling error.
It should be in all recent versions. I'm not sure about older versions. As for how else to do it, I'm not sure there is a way with Mplus, but you could do it with AMOS instead. I have many videos showing how to do this in AMOS.
Dr.Gaskin, whenever I try to run my CFA I get the error msg: " One or more variables have a variance greater than the maximum allowed...". Even though this is not true, my variables are item parcels with no missing value and with little variance. Any way to solve this problem?
That is odd. You can fix the variance to a specific value by using the @ sign. I would recommend setting to to something small like .05. Or you can set it to an inequality (e.g., >0 or
Hello Dr. Gaskin, How we can modify CFA command while some of our items have high variance? MPlus does not give me any result and the output shows a note saying some of items have very high variance therefore no result can be obtained. I asked some users and i was given some instructions. however, the software still does not give me any result and shows error. Can you help please? Thank you.
I'm a novice at Mplus, so I'm not sure. My initial guess would be to check the normality and outliers for whichever variables were causing the error. Then try to bring them to a normal distribution.
wow that helpful as usual, I just want to clarify the Likert scale is discrete or categorical variable? Simply with the Likert scale do I have to use STDY or STDYX ? sorry sir but am just a beginner in statistics want to learn. All the best for you.
thanks for the video. please, i want to ask about the normality in CFA. i have no normal data to do CFA in Amos, so i can't use " Maximum likelihood". also, i can't use "Asymptotic Distribution Free(ADF)" because my sample size is only (306). so my question is how can i do the CFA ? or, there is another way, how can i do the Satorra and Bentler scaling Correction for khy square in Mplus? for handle the problem of no normal data
Factor loadings are the extent to which an indicator "loads onto a factor", which is essentially the extent to which an indicator is correlated with all other indicators strongly loading on that factor.
Dear James, when tried to test 13 scales with 300 samplings, it didn't process on Amos. Would Mplus do the same or able to test such a case? Although Smartpls works in such a case, wondering whether it would be the case in Mplus.? Kind Regards...
13 scales (each with their own set of measures) or 13 measures/variables? Mplus should be able to process it either way, unless you are violating some assumptions.
This is considered a more rigorous approach, but it is rarely done because we often don't have enough sample size to split it without losing adequate statistical power.
hi sir, I'm Dwi Yan Nugraha, a psychology student at Makassar state university in Indonesia. I was very hampered because I did not have the Mplus full version software.
Hi James, I need to conduct second order factor structure on Mplus as my first order correlations came high. Do you have a video on this detailing the steps? And how to interpret the correlations from the second order structure? I’m struggling here. Thanks
I don’t have a video for that. However, here is some generic syntax that will guide you. DATA: FILE IS your_data.dat; VARIABLE: NAMES ARE Y1-Y9; USEVARIABLES ARE Y1-Y9; MODEL: F1 BY Y1-Y3; F2 BY Y4-Y6; F3 BY Y7-Y9; HOF BY F1 F2 F3; OUTPUT: STANDARDIZED;
Thanks a lot. I ran the syntax. Now, the higher order factor seems to have correlations of .956, .980, .892, .942 with the four lower order factors. Is that okay? There are four broken arrows with the four lower order factors (.086, .040, .204, .112). What do these mean? And how to justify the whole thing? Do we need to check the model fit again for this 2nd order structure? Because my RMSEA seems to have increased (.087) for this 2nd order analysis.
@@farzanaashraf4661 What do you mean by broken arrows? I've never seen that in Mplus. Model fit should be checked for any model used to test a hypothesis.
Dear James, what do you think about correlation levels on factors? for instance, I have a sclae and it has four factor, I conducted second order CFA. Correlation levels among the factors are .93-.95-.93-.90. Hovewer, CFI 948 and TLI 940. What is your suggestions? Thanks so much.
@@chios1976 Anything about higher order factors should be adequate as a citation. e.g., Sarstedt, M., Hair Jr, J. F., Cheah, J. H., Becker, J. M., & Ringle, C. M. (2019). How to specify, estimate, and validate higher-order constructs in PLS-SEM. Australasian Marketing Journal (AMJ), 27(3), 197-211.
@@GaskinationHi James, I need to conduct second order factor structure on Mplus as my first order correlations came high. Do you have a video on this detailing the steps? And how to interpret the correlations from the second order structure? I’m struggling here. Thanks
Hello doctor, my scale did not meet the fitting index after CFA validation, but the item's load is normal. I have analyzed all the errors and still cannot meet the requirements. What should I do?
Poor goodness of fit can be attributed to several common issues: a. Incorrect Factor Structure: The proposed model may not accurately reflect the underlying data structure. b. Omitted Variables: Important latent variables might be excluded, leading to misspecification. c. Incorrect Factor Loadings: The assumed relationships between observed variables and latent factors may be incorrect. d. Poorly Defined Indicators: Some indicators may not adequately measure the intended latent construct. e. High Measurement Error: Excessive error variance in the observed variables can distort the model fit. f. Insufficient Sample Size: A small sample size can lead to unstable estimates and poor model fit. g. Unrepresentative Sample: If the sample does not represent the population well, the model fit may suffer. h. Overfitting: An overly complex model with too many parameters can lead to poor fit. i. Underfitting: A too simplistic model that fails to capture the data's complexity can also result in poor fit. j. Non-Normality: If the data significantly deviate from normality, it can affect the goodness of fit. k. Missing Data: High levels of missing data or improper handling of missing data can degrade model fit. l. Outliers: Outliers can distort parameter estimates and lead to poor fit. m. Relying on a Single Index: Using only one fit index (e.g., Chi-square) can be misleading. It's important to consider multiple fit indices (e.g., RMSEA, CFI, TLI). n. Overly Restrictive Constraints: Imposing unnecessary constraints on parameters can lead to poor model fit.
Most people use an online survey service like qualtrics or surveymonkey. Once the survey is distributed, participants enter their responses and this automatically populates a database that can be downloaded as a .csv or .sav file.
good evening sir i am Krishnababu Sambaru working as a assitant professor at andhra pradesh in India i have a small doubt regarding spss and amos please do some help sir thanking you sir
A transcript? Click on the ... next to the share and save buttons below the video, then click on Open Transcript. If you mean the Mplus syntax, it is available here: statwiki.gaskination.com/index.php?title=Mplus
Hello Sir, In Amos, whenever i perform the CFA, and calculate the estimates, the window show that "The variable is represented by a rectangle in the path diagram, but it is not an observed variable". tell me why it is happened again again and i have observed that my data is correct.
Oh, you're in AMOS. Yes then, it is because of the wrong dataset (or maybe you didn't save that dataset before linking it), or a spelling error in the name of an observed variable (rectangle).
Thanks, James, You always do great!
You're great Dr. Gaskin.
Please also add in an elaborate video, how to report these analyses. Thanks a lot for sharing these videos. You truly are a beacon for novices like us
THANK YOUUUUUU SOOOO MUCHH
Hello Dr. Gaskin, Thank you for your video. What citation do you use to support the cut offs you use for goodness of fit estimates? (e.g. goof it is RMSEA 0.90, etc.) Thank you!
statwiki.gaskination.com/index.php?title=References#Model_Fit
Dr. Gaskin, thanks for the video, I am following the steps you have shown in that video for a measure that has four subscales. My output does not produce fit stats (RMSEA, chi square etc), can you advise on why it might be happening?
That is unexpected. I wish I knew Mplus better. I'm not sure why that would happen. I plan to learn Mplus better over the summer. Best of luck to you.
This is a very helpful video James. Have you figured out is there a simple way to extract CR, AVE and construct correlations from Mplus? I guess no excel stats package is available yet like with AMOS? Thanks for your help and keep up the good work!
I haven't figured that out yet. I got caught up in another major plugin project with AMOS. I'll look into Mplus again hopefully in August.
Thank you, it was very helpful! One question, do you know the reference you used to say that the correlations between factors should be above .8 or at least above .9? Can't find the .9 criteria!
Sorry about that imprecise language. I've learned more since making this video. Here is an updated one with more standard measures: ua-cam.com/video/4LGQ_t0xuj8/v-deo.html
@@Gaskination It was perfect, don't worry! Thank you very much!
The video was really helpful. I am just confused about STYX and STDY (which one is used for categorical var, as I got y for categorical and same for YX). plz clearify this.
STDYX and STDY just produce standardized estimates for the Y and X variables or just for the Y variables. Y is for endogenous and X is for exogenous.
Jeez James you are saving my ass here
Hello SIr,
I have used AMOS 21 where Confirmatory factor analysis was calculated where i got Chi-square = 528.338. Tell me what is the acceptable value of Chi-square.
chi-square is relative to the number of variables and the sample size. So, a complex model with a sample size of 500 may have a chi-square in the 10,000s, but a simple model with a sample size of 75 may have a chi-square in the 100s. The general rule is that you want to minimize chi-square. A good relative measure is the chi-square/degrees of freedom. This should range between 1 and 5. Better measures are the SRMR, CFI, and RMSEA though.
Thank you. It was so helpful! However, can I use Mplus on Macbook M1?
It should work through a virtual machine (like citrix). I'm not sure if it works directly on the new Mac. You'll need to reach out to the Mplus creators to check.
Great work indeed. Do we have to mention categorical variables in cfa input command? Also please make a more elaborate video on cfa
Generally, categorical variables do not belong in a CFA. Categorical variables only belong to formative factors, if at all.
@@Gaskination Yes but we are considering ordinal variables (from Likert scale survey questions) as categorical ...
@@mikahely6091 Mplus considers ordinal as categorical. Correct. SPSS considers ordinal as ordinal (not nominal)
Hello Sir,
whenever i perform the CFA, and calculate the estimates, the window show that "The variable is represented by a rectangle in the path diagram, but it is not an observed variable". tell me why it is happened again again and i have observed that my data is correct.
This means that the dataset you are linking does not include a variable with the exact same (case and spelling) as the variable you are trying to model. So either the dataset is incorrect, or the syntax includes a spelling error.
Hello Sir, is the diagram option available in all versions of MPLUS if not is there any other method to get the diagram?
It should be in all recent versions. I'm not sure about older versions. As for how else to do it, I'm not sure there is a way with Mplus, but you could do it with AMOS instead. I have many videos showing how to do this in AMOS.
Dr.Gaskin, whenever I try to run my CFA I get the error msg: " One or more variables have a variance greater than the maximum allowed...". Even though this is not true, my variables are item parcels with no missing value and with little variance. Any way to solve this problem?
That is odd. You can fix the variance to a specific value by using the @ sign. I would recommend setting to to something small like .05. Or you can set it to an inequality (e.g., >0 or
Hello Dr. Gaskin, How we can modify CFA command while some of our items have high variance? MPlus does not give me any result and the output shows a note saying some of items have very high variance therefore no result can be obtained. I asked some users and i was given some instructions. however, the software still does not give me any result and shows error. Can you help please? Thank you.
I'm a novice at Mplus, so I'm not sure. My initial guess would be to check the normality and outliers for whichever variables were causing the error. Then try to bring them to a normal distribution.
wow that helpful as usual, I just want to clarify the Likert scale is discrete or categorical variable?
Simply with the Likert scale do I have to use STDY or STDYX ?
sorry sir but am just a beginner in statistics want to learn. All the best for you.
Likert is considered categorical in Mplus (ordered categories). So, you can use STDYX.
@@Gaskination Thank you so much prof. Gaskin
thanks for the video. please, i want to ask about the normality in CFA. i have no normal data to do CFA in Amos, so i can't use " Maximum likelihood". also, i can't use "Asymptotic Distribution Free(ADF)" because my sample size is only (306). so my question is how can i do the CFA ? or, there is another way, how can i do the Satorra and Bentler scaling Correction for khy square in Mplus? for handle the problem of no normal data
You could try bayesian estimation in AMOS, or try a non-parametric approach like PLS.
@@Gaskination if you have any video about how we use bayesian in amos, please help me with it.
@@hibald9458 Unfortunately I do not...
ad tell me what is factor loading (lambda)
Factor loadings are the extent to which an indicator "loads onto a factor", which is essentially the extent to which an indicator is correlated with all other indicators strongly loading on that factor.
Hello, Can we calculate composite reliability and Cronbach alfa in Mplus? thank you.
I'm not sure. I would guess yes. However, I'm still a novice. I'll learn more this summer...
James Gaskin thank you. Can you please have a video on it if we can? Thank you for all of the great videos.
Dear James, when tried to test 13 scales with 300 samplings, it didn't process on Amos. Would Mplus do the same or able to test such a case? Although Smartpls works in such a case, wondering whether it would be the case in Mplus.? Kind Regards...
13 scales (each with their own set of measures) or 13 measures/variables? Mplus should be able to process it either way, unless you are violating some assumptions.
I have read that you must split 1/2 the data for EFA and then run CFA on the other 1/2 of data. Do you think this is necessary? Why or why not?
This is considered a more rigorous approach, but it is rarely done because we often don't have enough sample size to split it without losing adequate statistical power.
hi sir, I'm Dwi Yan Nugraha, a psychology student at Makassar state university in Indonesia. I was very hampered because I did not have the Mplus full version software.
I think they give out trial versions.
Hi James, I need to conduct second order factor structure on Mplus as my first order correlations came high. Do you have a video on this detailing the steps? And how to interpret the correlations from the second order structure? I’m struggling here. Thanks
I don’t have a video for that. However, here is some generic syntax that will guide you.
DATA:
FILE IS your_data.dat;
VARIABLE:
NAMES ARE Y1-Y9;
USEVARIABLES ARE Y1-Y9;
MODEL:
F1 BY Y1-Y3;
F2 BY Y4-Y6;
F3 BY Y7-Y9;
HOF BY F1 F2 F3;
OUTPUT:
STANDARDIZED;
Thanks a lot.
I ran the syntax. Now, the higher order factor seems to have correlations of .956, .980, .892, .942 with the four lower order factors. Is that okay? There are four broken arrows with the four lower order factors (.086, .040, .204, .112). What do these mean? And how to justify the whole thing?
Do we need to check the model fit again for this 2nd order structure? Because my RMSEA seems to have increased (.087) for this 2nd order analysis.
@@farzanaashraf4661 What do you mean by broken arrows? I've never seen that in Mplus. Model fit should be checked for any model used to test a hypothesis.
Dear James, what do you think about correlation levels on factors? for instance, I have a sclae and it has four factor, I conducted second order CFA. Correlation levels among the factors are .93-.95-.93-.90. Hovewer, CFI 948 and TLI 940. What is your suggestions? Thanks so much.
The first order dimensions of a reflective second order factor should be highly correlated. So, all these numbers make sense.
@@Gaskination Dear James, Thank you so much. Last thing, may you suggest a literature about your suggestion? Thanks again...
@@chios1976 Anything about higher order factors should be adequate as a citation. e.g., Sarstedt, M., Hair Jr, J. F., Cheah, J. H., Becker, J. M., & Ringle, C. M. (2019). How to specify, estimate, and validate higher-order constructs in PLS-SEM. Australasian Marketing Journal (AMJ), 27(3), 197-211.
@@Gaskination Thanks again...
@@GaskinationHi James, I need to conduct second order factor structure on Mplus as my first order correlations came high. Do you have a video on this detailing the steps? And how to interpret the correlations from the second order structure? I’m struggling here. Thanks
Hello doctor, my scale did not meet the fitting index after CFA validation, but the item's load is normal. I have analyzed all the errors and still cannot meet the requirements. What should I do?
Poor goodness of fit can be attributed to several common issues:
a. Incorrect Factor Structure: The proposed model may not accurately reflect the underlying data structure.
b. Omitted Variables: Important latent variables might be excluded, leading to misspecification.
c. Incorrect Factor Loadings: The assumed relationships between observed variables and latent factors may be incorrect.
d. Poorly Defined Indicators: Some indicators may not adequately measure the intended latent construct.
e. High Measurement Error: Excessive error variance in the observed variables can distort the model fit.
f. Insufficient Sample Size: A small sample size can lead to unstable estimates and poor model fit.
g. Unrepresentative Sample: If the sample does not represent the population well, the model fit may suffer.
h. Overfitting: An overly complex model with too many parameters can lead to poor fit.
i. Underfitting: A too simplistic model that fails to capture the data's complexity can also result in poor fit.
j. Non-Normality: If the data significantly deviate from normality, it can affect the goodness of fit.
k. Missing Data: High levels of missing data or improper handling of missing data can degrade model fit.
l. Outliers: Outliers can distort parameter estimates and lead to poor fit.
m. Relying on a Single Index: Using only one fit index (e.g., Chi-square) can be misleading. It's important to consider multiple fit indices (e.g., RMSEA, CFI, TLI).
n. Overly Restrictive Constraints: Imposing unnecessary constraints on parameters can lead to poor model fit.
Do you have a lecture on HLM in MPLUS, please?
not yet. I'm hoping to make many more Mplus videos this summer.
Plz make one on HLM kindly
how do we enter questionnaire in spss
and how do we know our questionnaire is correct and wrong
please clarify my doubts
thanking you sir
Most people use an online survey service like qualtrics or surveymonkey. Once the survey is distributed, participants enter their responses and this automatically populates a database that can be downloaded as a .csv or .sav file.
You probably don't care at this point but consider putting all the links to the videos in the series in the description of each video.
thanks for the recommendation!I have created a playlist with all these videos. Perhaps I'll add the link to the playlist in the description.
good evening sir
i am Krishnababu Sambaru
working as a assitant professor at andhra pradesh in India
i have a small doubt regarding spss and amos
please do some help sir
thanking you sir
Can I request the full version of the Mplus software that you are using
That would be illegal. I don't own it. I only license it..
can you share with me? Is there a Google Drive that I can go to
@@dwiyannugraha No. that would be illegal.
Can anyone please help me download Mplus 7 for mac os cracked version
That would be illegal.
Where is t-value in amos?
it is the critical ratio (C.R.). In the regression weights table, it is to the left of the p-value.
I need a script please
A transcript? Click on the ... next to the share and save buttons below the video, then click on Open Transcript. If you mean the Mplus syntax, it is available here: statwiki.gaskination.com/index.php?title=Mplus
Hello Sir,
In Amos, whenever i perform the CFA, and calculate the estimates, the window show that "The variable is represented by a rectangle in the path diagram, but it is not an observed variable". tell me why it is happened again again and i have observed that my data is correct.
Oh, you're in AMOS. Yes then, it is because of the wrong dataset (or maybe you didn't save that dataset before linking it), or a spelling error in the name of an observed variable (rectangle).