0
$\begingroup$

Could someone provide a sample interpretation for the following:

                                dependent variable
Categorial variable
 category 1                    Beta estimate of 1.78 (-3.39,6.69). 
 cat 2                         2.45 (-287, 9.34)
 cat 3                         4.49 (-3.65, 7.08)
$\endgroup$
2
$\begingroup$

It depends on whether you are using dummy coding or effect coding or some other coding. In dummy coding, a positive $\beta$ for a categorical variable means that the predicted value of the dependent variable is higher by $\beta$ when the independent variable takes that category than when it takes the reference category

$\endgroup$
2
$\begingroup$

Note that if the numbers in parens are the confidence intervals on the betas then for all of your coefficients the intervals include both positive, negative, and 0 values meaning that it is possible that there is no effect what soever or that it could even be negative and that the reason your are seeing positive betas could be explained by pure chance.

Before worrying too much about the size and direction of effects you should make sure that what you are seeing cannot be easily due to random chance (the sample you happened to get).

If you get meaningful estimates that are not due to chance, then see @PeterFlom's answer.

$\endgroup$
1
  • 2
    $\begingroup$ Yes, and @confusedstat you probably want also want a test for the overall usefulness of the categorical variable so you can test all categories simultaneously rather than just one at a time (which is all your current readout allows). This can be done via an F test comparing the model with the variable in to one without the variable - often a standard output from an ANOVA table based on the model. $\endgroup$ Mar 31 '13 at 3:07

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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