Statistical analysis of the effect of mailing letters

[amalali]

I want to analyse the result of completely randomized experiment of one factor (8 levels) with fixed effects. The purpose was to identify the effect of different letters mailed out to families in signing up to a children's program.

We had 7 samples each received one type of letter and we had a control group( with no letter).  I also want to determine the effect of letters by demographic characteristics (age, gender, language, geographical location).

I need help in carrying the statistical analysis  in SAS.

Prog_Status is the dependent variable holds the information if the person signed up or not due to the letter. It’s binary data with 0 ( no sign up) or 1 ( signed up)

Test variable is independent and holds the type of letter, 1 being control group and 2-7 being the type of letter recieved. 

I'd like to show the effect of letter by age categories (0-2, 3-6, 7-14), gender (F, M), Language (Englihs vs French) and by location (13 different locations)

I'd like to show also which letter is being most effect. 

I used the following but not sure if it’s correct or complete.

proc glm data=work.a;                                                                                                                  

class test;                                                                                                                            

model Prog_Status=test/solution;                                                                                                       

estimate 'control vs letter2' test 1 -1 0 0 0 0 0 0 / e;                                                                                

run;       

thanks,

[Doc_Duke]

You are on the right track, but you need to use PROC LOGISTIC because your outcome is binary.

There is a good BBU by Allison "Logistic Regression with SAS" that can help you through the specific comparisons you are looking for.

[amalali]

Thanks Doc_Duke. I'll have a look at the BBU.