06-11-2013 10:24 PM
I'd like to try the residual analysis after chi-square test.
But I can't find the procedure of residual analysis in SAS.
Do you know the procedure name of residual analysis?
Thank you for your help.
06-12-2013 12:49 PM
Please be a bit more precise. What sort of chi-squared test, and how did it come about? Ordinarily, I don't think of residuals AFTER a chi-square. I might use a chi-square to compare the cumulative distribution of the residuals to a theoretical distribution, but those residuals would come from some sort of linear model.
06-13-2013 09:40 AM
Thank you for your response.
The summarized table of 3 x 4 was analyzed by chi-square test.
There was significant differences among 3 groups as a result of the analysis, but we don't know which one has difference.
So I want to try residual analysis to know the significant difference of each cells.
06-13-2013 12:57 PM
You should get a copy of Agresti's Categorical Data Analysis. It goes over this in detail. Here are some of the key elements:
1. This is not residual analysis, as considered by most statisticians. It is subset analysis.
2. After running a full table and finding that there is a significant association between rows and columns, pick out the columns you wish to compare for a pairwise analysis. Here I assume rows are responses and columns are groups of some sort.
06-16-2013 09:58 PM
Thank you for your good advice.
I will read the copy recommended you.
Columns is groups of age such as 30-40,41-50,51-.
I hear pairwise (each other) analysis in more than 3 groups is prone to appear significant difference.
06-17-2013 08:46 AM
Yes, that is a consideration. To control this, you may wish to adjust the p values. Please look at The MULTTEST Procedure Example 61.5 Inputting Raw p-Values for an example of how to adjust for the comparisons you choose.