I ran an experiment with $n$ subjects. For each subject, I measured the correlation ($\rho$) between two variables. Thus, I ended up with $n$ correlation coefficients $\rho$. I wanted to test whether the population correlation coefficient is different from zero. What statistical test should I perform?
I had tried Fisher transformation on each subject's $\rho$, but not sure how to proceed.
One option is to use a simple ttest on the correlation coefficient after it has been transformed. I don't know if its best practice, but people use it commonly in neuroscience (e.g. in inter subject correlation).
Alternatively you can try to generate the null by randomizing in some way (e.g. shuffling one of the variables for each subject, calculating the correlation for each, then averaging across subjects). Repeating this many times will give you a distribution representing the null which you can compare against.