1
$\begingroup$

As I am testing a number of models, I notice that none of my demographic variables are ever signficant. For example, I am testing a model to predict the dependent variable 'perceived substitutability'. The independent variables are contentment about the content on traditional television and contentment about the content on online television. I run this model together with gender, age, professional status, income (yes/no), family composition and marital status. I used a hierarchical method where I used the demographic variables in the first block and the other variables in the second block. As I run the regression, all demographic variables were not significant (p>0,05). This is of course due to correlations between the independent variables.

If I use one demographic variable at a time to explain the dependent variable (so six times a simple regression where no other demographic variables are used), professional status and income become significant in the simple regression (not together because of course they correlate). Can I use this in my result-section: that in the simple regressions (1 dependent, 1 independent) they are significant while in the multiple they were not?

I am using SPSS

Improve this question
$\endgroup$
1
  • 1
    $\begingroup$ Then I would report the regression results from the two contentments as a model apart from the two demographic variables that were significant. (in a simple regression and not together) because if I use the contentments together with the demographics, the demographic variables become non-significant $\endgroup$
    – user45825
    May 21, 2014 at 14:32

1 Answer 1

Reset to default
1
$\begingroup$

You can certainly report both simple and multiple regressions in a paper; I see this done all the time. You can then explain the differences, just as you have done above - but instead of saying things like "of course" I would, in a paper, show that they are related.

I was a bit puzzled by income being (yes/no) but that may just be due to an odd choice of variable name.

Improve this answer
$\endgroup$
2
  • $\begingroup$ Ok thank you for your answer! Is this a good justification that I did simple regressions because of the high correlations between predictors? Because that is in fact the reason. When I put all of them together in a multiple regression they are not significant and therefore I cannot discuss demographics. I must add that the correlations are between ,200 and ,500 but still I assume this is the reason that together they are not significant while apart they are $\endgroup$
    – user45825
    May 21, 2014 at 16:04
  • $\begingroup$ Yeah, although you could also try collinearity diagnostics with condition indexes $\endgroup$
    – Peter Flom
    May 21, 2014 at 19:17

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.