04-12-2017 11:02 AM
I want to compare two percentages. I think Chi-Square test is appropriate for this. So, I have the following datasets;
Group Number of Success Total number
1 5 10
2 4 30
I need to compute the percantages of number of success in the first and second group and then compare them whether the difference between this comparision is statistically significant by using Chi-Square test. So the should be one more column:
Group Number of Success Total number Percentages
1 5 10 50% (5/10*100)
2 3 30 10% (3/30*100)
And then I want to see whether the difference between percantages (50% and 10%) is statistically significant by using Chi-Square test.
04-12-2017 01:48 PM
Chi Square would use the raw numbers from the 2x2 table, not the percentages.
To compare percentages you can use PROC TTEST but you need to make sure to include the correct N and possibly the standard deviation. You also need to factor in a continuity correction.
I recommend PROC FREQ instead of trying to test the percentages. The data lends itself to that format naturally.