# regression on age groups

#### sugabird22

##### New Member
Dear All,

I have the following question: Estimate a simple regression model of ‘libertarian-authoritarian’ attitude as a function of age groups and comment on the sign, significance and magnitude of the effects. I have one variable (and only one) called AgeGr and the age groups are 18-25, 26-35, 36-45, 46-55, 56-65 and 66+. How should i do a regression on this? Should i just do a regress libauth on AgeGr or should I separate the age groups such as having a separate dummy variable for every age group and then run a regression on every age group. (i.e. regress libauth on AgeGr18-25). So I would do 6 regressions. And also, how should i interpret the results? If i did the first way (without separation), should i interpret it such as increasing the age group by one (i.e from 18-25 to 26-35) leads to a xy change in libauth.

Cheers,

Andras

#### hlsmith

##### Omega Contributor
I am guessing you want to regress libert on the age groupings. The groupings should be dummy coded if your program does not automatically do that for you. Then one group would be the reference group the others are compared to it. Also if your outcome was libert (yes/no), you would want to use logistic regression.

#### sugabird22

##### New Member
Thank you very much. You helped a lot. I am guessing tabulating AgeGr would help dummy coding it. The lib-auth scale is from 1-5 so i can't use logistic regression though.

#### hlsmith

##### Omega Contributor
Perhaps look into ordinal regression.

#### sugabird22

##### New Member
Cheers! Also, just came to my mind, what should be meant by the magnitude of the effect in the question above?

#### hlsmith

##### Omega Contributor
Your magnitude of effect would be the odds.