Exploratory Factor*Analysis - Categorical & Ranked Variables - Post Estimation
Dear All, I am not from an English speaking background, so kindly excuse if I could not follow any rules of this forum.
I am working on primary survey data with 5000 households. I wanted to find 4-5 constructs containing various categorical and ranked variables within each component that explains certain decision behaviours between males and females. Please read my problems under this and try to answer whichever you can help me with.
.
First I used the -factortest- command on all the questions/ variables of interest. My KMO comes as 0.90 . Then I used -polychoric- on them and the error code of missing values came up although when i use -sum- command on the variables of choice, none of them has any missing value individually.
QUERY 1- Why is this happening? How to solve this error issue? And am I right in using the -polychoric- command for such variables (dummy all of them)?
.
Next, one by one I tried and then eliminated all such variables from this above exact exercise (even some important variables too) in order to find the final set of 19 variables on which running -polychoric- command does not yield any error message. And then my -factortest- command reduces the KMO value and gives me 0.745.
QUERY 2- Is it ok to remove those variables for which polychoric was going berserk and try like this? And is this 0.745 KMO value valid enough to run an EFA using polychoric after this?
.
Next, I rotated and used sortl command and then tried to run -alpha varlist,c- for the variables under each of the 5 factors with the highest variances but each of the 5 alphas gave me around 0.5 values.
QUERY- Is this cronbach alpha value acceptable?
General Query - Shall I proceed to CFA on the same model after this EFA although my KMO and Cronbach alpha values are this low? Where am I going wrong here? Are there any better ways of doing this whole exercise, from the top?
I am working on primary survey data with 5000 households. I wanted to find 4-5 constructs containing various categorical and ranked variables within each component that explains certain decision behaviours between males and females. Please read my problems under this and try to answer whichever you can help me with.
.
First I used the -factortest- command on all the questions/ variables of interest. My KMO comes as 0.90 . Then I used -polychoric- on them and the error code of missing values came up although when i use -sum- command on the variables of choice, none of them has any missing value individually.
QUERY 1- Why is this happening? How to solve this error issue? And am I right in using the -polychoric- command for such variables (dummy all of them)?
.
Next, one by one I tried and then eliminated all such variables from this above exact exercise (even some important variables too) in order to find the final set of 19 variables on which running -polychoric- command does not yield any error message. And then my -factortest- command reduces the KMO value and gives me 0.745.
QUERY 2- Is it ok to remove those variables for which polychoric was going berserk and try like this? And is this 0.745 KMO value valid enough to run an EFA using polychoric after this?
.
Next, I rotated and used sortl command and then tried to run -alpha varlist,c- for the variables under each of the 5 factors with the highest variances but each of the 5 alphas gave me around 0.5 values.
QUERY- Is this cronbach alpha value acceptable?
General Query - Shall I proceed to CFA on the same model after this EFA although my KMO and Cronbach alpha values are this low? Where am I going wrong here? Are there any better ways of doing this whole exercise, from the top?
Or should i go with mca?
Comment