I have monthly prevalence data (n=30-120/month) for the presence of a parasite (2x12 contingency table) that shows a significant chi square. I'm interested in comparing specific months to see if there are differences. Using the filter feature in contingency analysis I can quickly get multiple comparisons. However that is without any correction factor applied to control error rates. I have seen that JMP uses FDR to control for false positives but I can't find step-by-step information on how to use FDR to adjust my multiple comparisons.
First, is use of FDR a correct approach for multiple 2x2 comparisons (chi square)? My method to do multiple comparisons seems kind of kludgy (using the filter repeatedly). Second, is there a tutorial that shows how to apply FDR to a dataset?
I suggest 2 solutions.
1. Restructure your data table, so that each of the sub-analyses you want to do, have separate data values, and there is also a new column, that contains an identifier for each grouping. You can then run the Response Screening Platform indicating your X and Y columns, grouped by the new identifier column. That will give you your FDR.
Help==>Books==>Predictive and Specialized Modeling
2. Run Multiple Correspondance Analysis in the Consumer Research Analysis area. It will help you better understand the data and the different relationships.
You stated that you use a Data Filter multiple times to get your current analysis. Basically, what I am suggesting, is that you could create a separate grouping of data for each of the filterings you do, creating an identifying column that uniquely identifies each group, and then run the Response Screening. I envision that you would write a simple script to generate the groupings, and the Response Screening.
Here is a very simple example
Names Default To Here( 1 ); dt = Open( "$SAMPLE_DATA\hothand.jmp" ); Response Screening( Y( :First ), X( :Second ), Weight( :Count ), Grouping( :Player ) );