I have a datafile called 'original' and in it contains 4 variables, call them a, b, c, d, with n observations each. I then use proc surveyselect to draw 1000 resamples from the 'original' dataset with sample size n_b = n/4, the code is as follows:
proc sql noprint;
select ceil(count(*)/4) into :record_count
from original;
quit;
%put &record_count;
%let rep = 1000;
proc surveyselect data= original out=bootsample
seed = 1234 method = urs
sampsize=&record_count outhits rep = &rep;
run;
ods listing close;
This produces a datafile named 'bootsample' which contains 1000 samples with sample size n_b of each variable (a, b, c, and d) from the 'original' dataset. Each observation's replication ID is given by the variable "Replicate" (ranging from 1 to 1000).
What I need to do is this:Take Replicate = 1 (i.e., the first replication sample) and the variable a as an example. I want to calculate the following value of t: (if the picture below doesn't show, please see attachment of the picture titled "formula")
where mean(a) is the sample average of the variable a for replication sample 1, std(a) is the sample standard deviation of the variable a for replication sample 1, a_i represents each individual observation of the variable a for replication sample 1.
Then, I want to repeat the above procedure and calculate the value of t for all 1000 replication samples and all four variables: a, b, c, and d. I want to store the final result in a datafile called "result" that has 4 variables called a_t, b_t, c_t, and d_t (i.e., 4 columns) and the 1000 values of t of each variable in each row. So, graphically, a datafile structured like this: (if the picture below doesn't show, please see attachment of the picture titled "result")
Can anyone show me a template code that can achieve what I described above? I'm thinking maybe proc sql can do the trick, but I'm quite new to SAS and still don't really know the syntax very well. Thanks.
... View more
