Hi,
I was using the mean command to calculate confidence intervals and I found out that there is a difference in CI calculation:
#
cls
clear
version 14.1
input var group
15 1
19 1
17 1
13 1
12 2
18 2
11 2
19 2
end
mean var if group == 1
mean var, over(group)
#
Using the command with "if" I get a CI (-/+ 95%) of 11.89148 to 20.10852
Using the "over" command I get: 12.94728 to 19.05272
All calculations have the same standard error. As I have found out in Stata help is that "over" specifies that estimates are computed for multiple subpopulations, with special formula for subpopulation estimation. My question is: why would I need a special subpopulation calculation vor CI over groups?
-Nick
I was using the mean command to calculate confidence intervals and I found out that there is a difference in CI calculation:
#
cls
clear
version 14.1
input var group
15 1
19 1
17 1
13 1
12 2
18 2
11 2
19 2
end
mean var if group == 1
mean var, over(group)
#
Using the command with "if" I get a CI (-/+ 95%) of 11.89148 to 20.10852
Using the "over" command I get: 12.94728 to 19.05272
All calculations have the same standard error. As I have found out in Stata help is that "over" specifies that estimates are computed for multiple subpopulations, with special formula for subpopulation estimation. My question is: why would I need a special subpopulation calculation vor CI over groups?
-Nick