Multiple regression analysis or SEM

[katerina]

Hi there,

I want to find the effect of explanatory variables (factors) such age, job, sex on questionnaires' factors. My question is can I use multiple regression analysis on each factor, or I have to do SEM on three factors??

Thanks in advance

Katerina

[chiro]

Hey katerina.

Did you have a specific effect or a specific model in mind to test?

[katerina]

Hi Chiro,

I have a specific model to test.
Basically I got 3 DV correlated and 5 IV (age, job, sex as categorical and studies, days, and hourperday as continuous). and I want to test the effect of my IV on to DV. After those days I posted I decide first to do a MANOVA in order to test which of all my IV are significant and then a correlation of those who left on the model in order to check if there correlation among continuous and categorical and if there are correlation then I will use a MANCOVA model in order to test if the categorical var. is significant having in my mind the covariates.........I am so confused!! Are all of this correct or I have to change my plans???

Thanks in advance!!!
Katerina

[chiro]

Do you have a specific model though you want to test?

Typically if we want to test a model we write something like say Y = B0 + B1*X or Y = B0 + B1*e^(X) or something along those lines. Do you have such a model description?

Normal regression modelling is one way of fitting data to a specific model and the standard errors amongst other things can tell you how well the fit is for that model: it's not the only way but it is one way.

[Lazar]

SEM would be a relatively efficent way of doing what you want as everything can be fitted in a single model. Under most conditions the point estimates of a SEM and seperate multiple regressions should be very similar.

[statisticsdoc]

I would also favor SEM for the sake of having an integrative model, even more so if your continuous DVs are fallible scores from multi-item inventories that contain some element of measurement error that you can include in the model.