Obsidian | Level 7

## How to compare estimations from category level to the one in total

Hello,

I am studying a binary variable A (Yes/No) across different groups. e.g. immigrant, agegroup, provinces...... I use the syntax below to estimate the p-value for proportions of non-immigrant vs that of immigrant (the syntax is for SUDAAN)

proc descript data=mydata design=BRR;

Weight pweight;

repweight BSW_001-BSW_500;

var A  A ;

Catlevel   1 2;

class immigrant;

contrast immigrant= (-1 1)/ name= "non-immigrant va immigrant";run;

for the above one, it is easy because i can consider any group as the reference group. But, when it comes to multicategory groups like agegroup it could be something like this:

proc descript data=mydata design=BRR;

Weight pweight;

repweight BSW_001-BSW_500;

var A  A  A;

Catlevel   1 2 3;

class agegroup;

contrast agegroup= (-1 1 0)/ name= "46-55 vs 20-45";

contrast agegroup= (-1 0 1)/ name= 56+ vs 20-45";

run;

The syntax above will estimate the p-value for each age group of 46-55 or 56+ vs 20-45. But, I would like to estimate the p-value for each age group vs total population.

Thank you.

2 REPLIES 2
Super User

## Re: How to compare estimations from category level to the one in total

Unfortunately that looks like SAS callable SUDAAN code. Proc Descript is not a SAS product so you may not get a good answer here quickly. You might be better off asking on a SUDAAN board. It has been about 13 years since I had access to SUDAAN so can't make a good suggestion.

Obsidian | Level 7

## Re: How to compare estimations from category level to the one in total

Thank you so much for your response. Yes, it is SAS callable SUDAAN code.
Is there a SUDAAN board here or any other communities that you may know?

Discussion stats
• 2 replies
• 383 views
• 0 likes
• 2 in conversation