This FAQ is an edited version of a question and answer that appeared on Statalist.
|Title||Accumulating results from immediate commands|
|Author||Nicholas J. Cox, Durham University, UK|
|Date||March 1999; minor revision March 2001|
I am trying to use Stata to calculate confidence intervals quickly for a large amount of data. I have been using the immediate command cii to calculate each confidence interval, but I do not want to have to retype the results to make use of them. How do I accumulate the results of each calculation automatically into a new dataset?
You are using Stata as a calculator, typing
cii 12 56 34 cii 21 65 43
and so on, where the three numbers are the number of observations, the mean, and the standard deviation in each case.
To accumulate the results, we exploit the fact that cii leaves in its wake not just the printed results but also saved results that can be used either interactively or in a program. We can pick those up and put them in variables as part of a dataset that grows as we calculate.
First, set up the scenery. If you have data in memory, clear the data and type
set obs 1 gen N = . gen mean = . gen se = . gen lb = . gen ub = .
Then set up a do-file, for example, mycii.do:
-------------- mycii.do noi cii `1' `2' `3' qui replace obs = r(N) in l qui replace mean = r(mean) in l qui replace se = r(se) in l qui replace lb = r(lb) in l qui replace ub = r(ub) in l local n = _N + 1 qui set obs `n' -------------------
The l in the code above, in l, is the letter l (standing for last), not the numeral 1 (which would mean first). In this program, the r( ) are the saved results documented in [R] ci. The `1', `2', and `3' refer to the three numbers supplied to cii, its arguments in programming jargon.
run mycii 12 56 34 run mycii 21 65 43
Each time you run this do-file, the last observation (initially also the first) will be replaced, and the number of observations in the dataset will be bumped up by 1.
You can promote your do-file to a program:
-------------- mycii.ado program def mycii version 7.0 cii `1' `2' `3' qui replace N = r(N) in l qui replace mean = r(mean) in l qui replace se = r(se) in l qui replace lb = r(lb) in l qui replace ub = r(ub) in l local n = _N + 1 qui set obs `n' end -------------------
mycii 12 56 34
After the last calculation, you have a new dataset. Delete the last observation, which is all missing values.
The same approach will work with any immediate command. Just write your do-file or program to pick up the saved results as documented in the manual entry on the immediate command.