Does a significance test on Pearson's correlation coefficient require the dataset to be iid or at least independent?

[Alice Campbell]

I am trying to conduct a simple correlation significance test and realized that some data points might be dependent. I have been reading up on wikipedia (https://en.wikipedia.org/wiki/Pearso...t-distribution) and it doesn't mention that the underlying dataset need to be independent. Feels something is wrong here as the data could be driven by random numbers of correlated data ...

[Joro Kolev]

I think the significance test that Stata computes in -pwcorr- depends on the iid assumption.

But you can easily go around this. Just calculate the correlation through regression, and then you have testing robust to heteroskedasticity and auto correlation.

[Alice Campbell]

thanks! that's super helpful