Validity Tutorial

This video walks you through how to establish convergent and discriminant validity by calculating the MSV, ASV, CR, and AVE for complex structural models in AMOS, using my handy dandy stats tools package.

Category:

Education

Tags:

• Statistics
• Validity
• CFA
• Confirmatory Factor Analysis

[czerkhan]

the maximum items that included in the Statistics sheet is 14. if there are more than 14 items than how can modify the exel sheet from more than 14 items...

[Gaskination]

@czerkhan Wow! more than 14 must take your computer eternity to calculate. If it is more than 14 on a single latent variable, then I would argue that you should cut back. If it is more than 14 latent variables, then you should still probably cut back and seek for theoretical parsimony. All of this is moot anyway, because the current version of the stats tools package makes this version totally obsolete. Just use the new version and it will take care of it for you, even if you have 100 variables!

[mhdamro]

Hi .. i just have one question ... by computing cronbach alpha for 10 constructs in spss, i have got different CR values from the same set of data using the approach in this video (AMOS + your excel sheet) ...!!!! actually 9 constructs were the same ... but in one construct i got Cronbach alpha in spss of 0.42 and CR 0.72 in your excel sheet .... i wonder if you know why ? thanx in advance

[Gaskination]

@mhdamro

Are you looking at Composite Reliability in SPSS, or Critical Ratio, or Cronbach's Alpha? The algorithms for Composite Reliability and Chronbach's Alpha are different, and Critical Ratio is a very different algorithm. Also, Chronbach's Alpha is very sensitive to the number of items per factor.

[mhdamro]

check your googlemail ..

[Gaskination]

@mhdamro

This baffle me. The code is not very robust. It may be something as silly as the name of the factors. What are the labels of the factors? If they are something like "CR" "AVE" "ASV" or "MSV" or "e12" etc., then their might be problems.

[mhdamro]

@Gaskination on SPSS i am looking at cronbach alpha under "reliability statistics"

[mhdamro]

@Gaskination i have used the new tool.... SPSS Cronbach = 0.42 , old Excel tool= 0.72 , new excel tool = 0.72. on other construct i have SPSS Cronbach = 0.14 , Old Excel tool = 0.65, new excel tool = 0.513 ... the later construct of 0.14 reliability, i am sure its very unreliable from observing the items responses ... but why on 2 constructs i get the inconstant output while the others are ok ...?

[mhdamro]

@Gaskination i have used the new one as in your other video .... still the same problem ... on your tool it gives CR 0.721 on the SPSS it gives 0.424 ... i should not that the other constructs are ok except an additional construct where on new excel tool is 0.513, old excel tool is 0.65 SPSS is 0.14, !!! i tend to think the SPSS is true because i can see that this construct is unreliable by looking at the item responses . am left with one question, why some constructs are OK while 1 or 2 not ?

[Gaskination]

@mhdamro

Wow! That is very different. I suggest using my new tool and new method. See the "Validity during CFA made easy" video. Usually CR and CA are very similar (if not identical).