Read page 11 of chapter 13 of the ActEd notes.
Two points to remember:
1. The t-test statistic only holds under \(\rho =0\). In other words, the given statistic will follow a t-distribution only when you assume \(\rho =0\). So, if you are doing a hypothesis test with Ho: \(\rho =0\), you can use it.
2. The Fisher test statistic works for any value of \(\rho\) (including \(\rho =0\)). In other words, the statistic will follow a Normal distribution whatever be the value of \(\rho\). Hence, this statistic can also be used to derive a Confidence Interval for \(\rho\).
Check Q11 of IAI May 2014 paper as well.
Last edited: Oct 26, 2014