Help using Base SAS procedures

proc freq weight

Accepted Solution Solved
Reply
Occasional Contributor
Posts: 17
Accepted Solution

proc freq weight

Excuse my ignorance I’m new to SAS. Anyone have a suggestion on how to analyze
the following. The data consists of two data sets one being a control the other
representing implementation. A program was implemented to potentially increase attendance
in local schools. So each data set represents a list of schools and the attendance rate for each school. I took a
categorical approach to the data and used proc freq including the weight
command in order to generate a chi-sqre. I have a feeling my approach is incorrect


Accepted Solutions
Solution
‎09-25-2015 06:23 AM
Super User
Posts: 11,343

Re: proc freq weight

In a pedantic scholarly mode I would say that the analysis plan, the types of tests to be conducted with the data should have been decided upon before the data was collected.

The main thing to consider is what are you looking to compare? If it is a rate, then t-test for the mean rate between groups could be likely.

Chi-square would tell you if the distribution of responses was similar. Which works much better with categories than the almost certain different rate for each school.

I would recommend structuring your data into a single data set with a variable to indicate source, control or test, and start with tests of normality to see if t-test or other approach is needed, likely if the sample number of schools is small.

Basic way to combine the data:

data combined;

     set

          ControlData (in=incontrol)

          TestData

     ;

     if incontrol then source='Control';

     else source='Test';

run;

The Source variable could then be used as a grouping or class variable in many procedures.

I hope that the weight variable is the basic number of enrolled children in the school.

If the schools represent different populations, such as elementary, middle / junior high, high school it might be helpful to include that as a category as you may different results between the grades.

View solution in original post


All Replies
Solution
‎09-25-2015 06:23 AM
Super User
Posts: 11,343

Re: proc freq weight

In a pedantic scholarly mode I would say that the analysis plan, the types of tests to be conducted with the data should have been decided upon before the data was collected.

The main thing to consider is what are you looking to compare? If it is a rate, then t-test for the mean rate between groups could be likely.

Chi-square would tell you if the distribution of responses was similar. Which works much better with categories than the almost certain different rate for each school.

I would recommend structuring your data into a single data set with a variable to indicate source, control or test, and start with tests of normality to see if t-test or other approach is needed, likely if the sample number of schools is small.

Basic way to combine the data:

data combined;

     set

          ControlData (in=incontrol)

          TestData

     ;

     if incontrol then source='Control';

     else source='Test';

run;

The Source variable could then be used as a grouping or class variable in many procedures.

I hope that the weight variable is the basic number of enrolled children in the school.

If the schools represent different populations, such as elementary, middle / junior high, high school it might be helpful to include that as a category as you may different results between the grades.

Occasional Contributor
Posts: 17

Re: proc freq weight

Thank you ballardw for your response. The analysis plan was decided prior to collection but I am an intern and am detached from that process. I found myself questioning the proposed approach and will just leave it at that. Thank you again for your help!

Super User
Posts: 11,343

Re: proc freq weight

It is not uncommon that something in the results from the original plan raises questions. Sometimes they are additional interesting results, sometimes they are flaws in the data collection. So don't be afraid to experiment, but start with the plan.

Other ways to slice the data for comparisons could be by some category of school size (total enrollment), urban/suburban/rural locations, if you can get a good poverty index (School Free and Reduced Lunch participation rates may be available), ethic make-up.

Of course slicing the data more ways simultaneously requires more sample so may not be practical.

Occasional Contributor
Posts: 17

Re: proc freq weight

Thank you again ballardw!

🔒 This topic is solved and locked.

Need further help from the community? Please ask a new question.

Discussion stats
  • 4 replies
  • 390 views
  • 2 likes
  • 2 in conversation