01-16-2015 10:16 PM
I want to know the procedure(s) which will best compare and contrast the the performance of an event before and after implementation of new policy or after certain date?
For example: In a industry they have implemented new policy, so they wanted to compare the effectiveness of the policy, by comparing the before and after performance for one year, since the policy was implemented, they would like to compare the outcome (Percent, significance testing) based upon certain event?
Please let me know what procedures will help in doing these type of analysis?
01-16-2015 11:13 PM
Do you have a control group to test against? How can attribute changes to implementation versus other factors?
You haven't provided enough information but it sounds like a basic ANOVA or even t-test procedure.
01-19-2015 09:45 PM
Thanks for the reply. Actually My data is the from the meet industry where they perform sampling for few pathogens, so to prevent food borne disease they have implemented new inspection procedures for the industry. so now they have to compare the rates of positive counts of pathogens, their significance testing using certain model and volume weighted positive rates (depends on the amount of meet produced in that industry) before and after the implementation of the inspection for a certain period of time may be months, or quarter or bi-yearly, yearly and so on. so now they wanted to calculate this comparison in one procedure. I wanted one procedure which could do it.? if you have worked on this type of scenarios, please do help me?
01-20-2015 12:00 AM
You may want to control for different plants, meats, etc so you may going into proc panel or mixed here.
01-20-2015 10:46 AM
Rates are typically modeled using Poisson or negative binomial regression. This can be done in PROC GENMOD. See the discussion and example in this note.
01-19-2015 10:54 AM
I am not sure you have access to the SAS/ETS tools but this is exactly what PROC PANEL is built for. Here is a link to the documentation SAS/ETS(R) 13.2 User's Guide
It certainly is possible to do this with something like PROC MIXED but it would take quite a bit more coding to accomplish.
01-19-2015 03:32 PM
And if it is a very simple before/after comparison on a single variable, PROC TTEST may be all that is needed.
01-22-2015 11:54 AM
I used proc mixed, it gives the output more similar to the required output. Thanks to everyone for the reply and help.