Hi, I am trying to do a Monte Carlo simulation of ANCOVA. I generated random variables and simulated data from regressions models below. x is a covariate, g is a grouping variable (0 for a control group, 1 for a treatment group), y is a dependent variable. I am manipulating coefficients. In this syntax, I set the treatment effect on y to be 0, the treatment effect on x to be 0.6, and the effect of x on y to be 1. I also set the number of samples to be 1000 and the sample size to be 800 (about a half being a treatment group). data MC (drop=i);
call streaminit(12345678);
do sampleID = 1 to 1000;
do i = 1 to 800;
g = rand("Bernoulli", 0.5);
x = 0.6*g + rand("Normal");
y = 1*x + 0*g + rand("Normal");
output;
end;
end;
run; So, after that, I need to test the effect of g on y in each of 1000 samples. I wrote this syntax: proc glm data = MC;
BY sampleID;
class g;
model y = g;
run; It works but the problem is that I need only p values (for F values) from the output to count how many were less then .05 (which will serve as an estimate of Type I error rate). So, I can't figure out how to put these p-values in a data set, an array, or something like that (extracting p-values by hand for 1000 samples is not an option...). I would very much appreciate any help and/or feedback. Thank you.
... View more
