Mean Differences: Continuous IV and Categorical DV

[Yawo Kokuvi]

hello,

I am doing a bivariate analysis looking at the relationship between type of health care sought (none, routine, emergency) by many socio-demographic variables - age, income, gender, race, etc.

I am particularly interested in calculating mean age or income for those who seek each kind of health care, and to see if these mean differences are statistically significant.

I know T-test or Anova can be used, but only if the DV is continuous. In this case, the DVs are categorical and IVs include categorical and continuous variables. Crosstabs can handle the relationship between categorical IVs and categorical DVs.

How about a relationship between continuous IV and categorical DVs. Is bivariate regression the only option available. ?

thanks - cY

[Carlo Lazzaro]

Yawoo:
if you mention anova as a statistical tool for comparing the means of >2 groups, you are implicitkly sponsoring linear regression.
If your dependent variable is categorical with two or more non-ordered levels, you can go -logit- or -mlogit-, respectively.
All that said, I usually find the so called bivariate regression (that is, a series of simple regressions) quite unsatisfactory (although often reported in technical journal), because they do not consider the role of the other predictors in determining the contribution to variation of the conditional mean given by the independent variable of interest.

[Yawo Kokuvi]

thanks Carlo. the bivariate is actually a precursor to the multivariate, which is the main analysis. so essentially, you are saying that with categorical variables, logit or mlogit is the right approach, right?

thanks - cY

[Carlo Lazzaro]

Yawo:
you're correct.
If you have dependent, non-ordered categorical variables, depending on the number of levels (2 or more), -logit- or -mlogit-, respectively, are the right approaches.

[Yawo Kokuvi]

thanks very much - cY