SPSS..model 1 or 2?

[GetItDone]

I`m doing multiple regression in SPSS with Enter-method. The output i use is Model summary, ANOVA, and Coefficients.

I`ve put in some control variables/covariates , and SPSS shows this as model 1 and model 2 with and without the control variables.

But which of these models should i choose to report?
Is it model 1, where model 2 is just to show how the control variables effect the independent variable(s)?
That is at least what i`ve done for now. But i`m pondering if i should choose model 2 if the R-square of the model is higher than model 1? Or if the control variable is significant?


I haven`t found anything on google or guides to point me in any direction regarding this...

[Dr.Appalayya]

Choose the model whose R-Square is higher.

[GetItDone]

Choose the model whose R-Square is higher.

Thank you very much. I suspected it, though it was either the Beta-coefficients or the R-square i thought about as decisive..and if the control variables were significant thay might have to be taken into account. But i`ll go with the easy and sensible "R-square" alternative. ;-)

[Link]

Choose the model whose R-Square is higher.

Be careful when doing that. The model with more parameters will always have a higher R-square value. Does SPSS have an "adjusted" R-square value? Some statistical software will calculate that to try and account for the number of parameters in the model.

[GetItDone]

Be careful when doing that. The model with more parameters will always have a higher R-square value. Does SPSS have an "adjusted" R-square value? Some statistical software will calculate that to try and account for the number of parameters in the model.

Yes, SPSS have the adjusted R-square for the model.
Example (Roughly edited from SPSS table... Model 2 consists of independent/dependent and control variables):

Model Summary(c)
Model R R Square Adjusted R Square Std. Error of the Estimate
1 R =,481(a) R-Square= ,232 Adjusted R Square= ,225 Std.error of the estimate= ,99145

2 R= ,487(b) R Square= ,237 Adjusted R Square= ,215 Std. error of the estmate= ,99740

a Predictors: (Constant), Market orientation
b Predictors: (Constant), Market orientation, Belongs to a hotel chain, Number of employees in the hotel
c Dependent Variable: Information Quality

or....

Model Summary(c)
Model R R Square Adjusted R Square Std. Error of the Estimate
1 Model R,569(a) R Square= ,323 Adjusted R Square= ,310 Std. Error of the estimate= ,89136
2 Model R,582(b) R Square= ,339 Adjusted R Square= ,320 Std. Error of the estimate= ,88512

a Predictors: (Constant), AnalyticalCRM, Information Quality
b Predictors: (Constant), AnalyticalCRM, Information Quality, Number of employees in the hotel
c Dependent Variable: Marketing Innovation




I`ve seen my peers using R Square..not Adjusted R Square. But i`ve also seen Adjusted R Square being used in some tutorial (not related to my studies). I do not use a lot of variables in the model(s). SPSS can only run one dependent variable at the time, and i have two control variables (at most) along with max two other independent variables.

[Link]

I'm not sure what you're trying to do between the top two and the bottom two models. However, looking at the top two models, it looks like model 2 is a better fit.

I'm not surprised that a lot of your peers use r-square. Its very straight forward and easy to understand. However, there still should be caution when using it.

You should take a step back though and think about what it is that you're really trying to answer. What is your question? My co-worker/mentor always tells me that I should have that in the front of my mind, and that the question should then guide how I set up my models. Any fool can throw in variables and get results. Good models require lots of thought though.

HTH

[WeeG]

I would try running this model via the generalized linear model (SPSS got it), and then you can compare the models using AIC and BIC