Confirmatory factor analysis using AMOS (2017)

[@MaherGeorgesElmashhara]

The best video on KZbin that explains the CFA using AMOS.

[@mikecrowson2462]

DEAR VIEWERS: I A NEW VIDEO ON CONFIRMATORY FACTOR ANALYSIS (CFA) WITH AMOS THAT IS HOT OFF THE PRESS (AS OF SEPT 2020). THE VIDEO IS A BIT HIGHER QUALITY AND ALSO CONTAINS A POWERPOINT AND DATA YOU CAN DOWNLOAD FOR FREE UNDERNEATH THE VIDEO DESCRIPTION). PLEASE CONSIDER CHECKING IT OUT AT @hDpA (Confirmatory factor analysis in AMOS with Personality Disorder Scale items (Sept 2020))

[@vanessad5269]

Thank you! I want you to know that even after two years your valuable teaching and sharing of knowledge is benefitting the world and was a massive help to me. I will forever appreciate this. Thank you!

[@User255tv]

Hi Crowson.Thank you for this demonstration. Be blessed for sharing this video. I believe millions of people will find it useful. God bless you and continue sharing more videos.

[@mikecrowson2462]

Hey, thanks for the kind remarks. Much appreciated!

[@JamesonGoto]

A great video. It made my learning easy

[@kudratkhuda1372]

great video. liked the logical flow of discussion

[@karaklove]

Hello, thank you for this video, I want to know if I can do CFA for (IV, DV, Med and Mod) together? where the IV has 5 dimensions and each dimension has 3 or more items, while the DV, Med and Mod have only items (9, 13, 4) respectively.

[@nyarkobenjamin3104]

great video But how do i deal with my moderator variables after this

[@zhavitle]

Very helpful and user friendly video. thank you. how would you advice to handle the low covariance between the model's factor's in cases where my model predicts a significant correlation between the three factors, but the correlations are not found to be significant?

[@mikecrowson2462]

Hi. If two of the correlations not significantly correlated, then you might treat them as orthogonal in a separate factor analytic model and see whether the fit from the model involving the 3 correlated factors decreases significantly (using a chi-square difference test) as compared to a model in which you do not include a correlation (double-headed arrow) between the uncorrelated factors. I expect the fit won't change significantly, and if that's the case, the preferred model would be the simpler one (i.e., the one not including a correlation between factors). Hope this helps.

[@nab3230]

Hi.. i wanna ask y cant i open the data variable? I already add the data from the spss

[@tanyat4533]

Hi may I ask if both independent and dependent variables can be added to cfa? Or just independent variables are added?

[@chesskaboomgustz3755]

CFA can be done including both IVs and DVs in one model. So there will be two part of that CFA. One will be measurement model where relationship between items and factors will be tested and other will be structural model where path analysis will be accomplished to test the relationship between IVs and DVs. The video we just say does not include any IV or DV rather it confirms the factor structure of some concepts which is done after EFA. These concepts will become IVs and DVs in full CFA model i.e. measurement+structural model

[@hadiyasrebdoost4360]

Hello dear mike I want to know that can we do CFA with bootstrap and how? Thanks

[@varshagautam-w2f]

I need a help regarding CFA

[@runaheydahud6017]

Hey absolute great video. I have one question though. I have 4 constructs influencing perceived risk and then that one to influence purchase intention. When i wanna do the analysis it says "the following variables are endogenous but have no residual (error) variables": It allows me to continue anyways but when i have the actual data, I cant click on standardized estimated so they neither show up in the graph nor in the data table. Any ideas ?

[@mikecrowson2462]

Hi there. It sounds like you are specifying a path analysis with latent variables , with certain latent variables predicting others. Each variable with one or more arrows pointing to it is an endogenous variable in the model and each endogenous variable requires an error term. That includes models where you have latent variables predicted by others. See my other videos on path analysis with latent variables for more details!

[@sandeep202006]

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.

[@mikecrowson2462]

Hi Tourism, the only thing I can think of is the name you provided in the variable name box (under object properties) is not named the same thing in your SPSS dataset you imported into AMOS. The variable in the Variable Name box MUST match the variable in your SPSS file. If you want the diagram to show a different variable name, you use the Variable label box to type that in (but the name of your observed variable must match that in the SPSS file. Hope this helps.

[@mengzhenzhang8590]

Hello Sir, I have the same problem like Tourism. And I have checked the variable names in both SPSS and Variable Name Box for several times, they are exactly the same. And I'm sure they are observed variables... I wonder, what other reasons of it could you figure. Thank you a lot!

[@mengzhenzhang8590]

Hi i've met the same case as yours, have you solved it and may I know how you made it? Thanks a lot!

[@robinnelson1455]

When this error popped up for me, I had typed my variable name in the variable label box, not the variable name box.

[@omerrr09]

I could not understand why the regression coefficient was entered arbitrarily. In the lisrel ı did not see such a thing if I remember truely?

[@mikecrowson2462]

The fixed coefficient of 1 is necessary to identify the model. Basically, the latent variables have to have a measurement scale assigned. One option is to scale each latent factor in relation to one of the indicator variables associated with that factor. A different option is to fix the variance of the latent variable to 1. In AMOS the later will be apparent if you request the standardized solution, whereas the fixed loadings of 1 show up under the unstandardized solution. If you do not address the identifiability problem, then the model cannot be estimated. Cheers!

[@omerrr09]

@@mikecrowson2462 Thank you so much for fast and clear answer.

[@malisahlatip282]

Hi. Your video are very helpful. I want to ask. I have eleven factors to be studied. And a total of 76 items. I wonder how i want to draw it on the template. The template looks quite small. Please help me. Thank you in advance

[@mikecrowson2462]

Hi. You can click on the View tab and select (dropdown) Interface properties. Under Paper size (under the Page layout tab) and click on Landscape. This will give you a lot more area to work from. You have a lot of variables you are planning to use, so you may have to draw everything smaller using the drawing icons (the rectangles, circles, and arrow).. To keep everything an appropriate size, you might draw out a one factor model fairly small, and then click on select all items and then copy that section for other portions of the factor model. Might save you some time and headache trying to reshape each individual portion of your model. (i've done this plenty of times).

[@malisahlatip282]

H. Michael Crowson Hi sir, I want to ask question. I hve run Cfa for my data. I have 12 factors. With total of 76 items. And i have 367 samples. However, the gfi is around 0.85 after i delete the items with loading below 0.6, and items in MI. I also had remove the outliers, however it made the gfi decrease. Fyi, the CMIN, CFI, TLI, and RMSEA, all show acceptable results. I wonder if GFI 0.85 is acceptable?

[@mikecrowson2462]

Hi there. In general, GFI's less than .90 are not preferable. Values =>.95 suggest a good fitting model; although values in the .90's seem to still be regarded as indicative of acceptable fit. Even so, I probably wouldn't spend my time chasing after an acceptable GFI by modifying the model repeatedly. You have to look at the fit indices as providing different lenses through which to evaluate fit, and they won't all necessarily agree. The ones that seem to currently be given more weight when evaluating fit are the RMSEA and CFI.

[@vatsalpriyadarshi1256]

I am sorry to say even though your explanation is very good but there seems to be glitch in your capturing software because there's a delay in what you say and what is been shown in the video. For example play through play through 19 min part. When you said this you meant something that you were pointing to in the descriptive statistics part. But the capturing software displayed that after delay of few seconds. This caused confusion as to exactly what descriptive statistics for the goodness of fit you were pointing to.

[@yuwu22]

kzbin.info/www/bejne/hZWlen2ofNSplZI start talking about the model fit statistics