Regression with latent variable

[Swede159]

Hi,

I'm doing some research for my dissertation and need to construct a latent variable that includes observations of various economic determinants (IVs).

I'm using SPSS, but I don't have access to AMOS, and a additive composite index lacks reliability (Cronbach's Alpha is too small). Does anyone have any suggestion of how I can construct a composite index that reflects different weightings of different measures (observations) that can be made by SPSS without AMOS? Perhaps a way of weighing observations by regression coefficients from a multiple regression analysis of individual IVs?

Thanks

[lumhearts]

You can do that with exploratory factor analysis in SPSS. I'm not sure it will give better results than the additive method - but maybe. Just click on the scores button and save them.

[spunky]

Hi,
I'm using SPSS, but I don't have access to AMOS

i'm not sure whether this would help or not, but have you considerd using R?

they have quite a few excellent packages for latent variable modelling (i mostly use ltm and lavaan) which do the same thing as AMOS... with the added benefit of them being free and easily downloadable, of course.

[Swede159]

i'm not sure whether this would help or not, but have you considerd using R?

they have quite a few excellent packages for latent variable modelling (i mostly use ltm and lavaan) which do the same thing as AMOS... with the added benefit of them being free and easily downloadable, of course.

Thanks, I've checked it out and it looks like it does exactly what I need to do. I've hit a bit of a wall, however. I've specified a model with latent variables and a regression - but I get an error when I try to fit it to the data, using SEM. If you're familiar with the program (I've only had today to look at and try to figure it out), maybe you could help. I have specified the following model:


> Comp.model <- '
+ # latent variable definitions
+ ec =~ Employees + MCap + Countries + BU + Schooling + EXP + Tenure + ROSF + ROA + PER
+ BOCec =~ BOC * ec
+ # regressions
+ W ~ BOC + ec + BOCec + Position + Year + Hotel + Foodservice + Passenger + Beverages
+ # residual correlations
+ Employees ~~ MCap
+ Countries ~~ BU
+ Schooling ~~ EXP + Tenure
+ ROSF ~~ ROA
+ '

And have tried to fit it using:

> fit <- sem(Comp.model, data=ExecutiveYears.sav, missing=”ml”, estimator = ”wls”)

I then get an error message saying 'Error: is.data.frame(data) is not TRUE.'

Do you have any thoughts on what this might mean and how I can fix it?

Again, thanks a lot for your suggestion.