Suppose I obtain a coefficient $\hat{\beta_i}$ and it has 0.01 < p < 0.05. I believe the effect is real so I want to gather more data. Assuming $\hat{\beta_i}$ is close to the correct coefficient, how many more data points are required to obtain p < 0.01?

  • 1
    $\begingroup$ A p-value is a random variable, so you can at best calculate the probability Pr(P<0.01). Given the distribution of $\hat{\beta}_i$ this shouldn't be too hard. $\endgroup$
    – Knarpie
    Nov 21, 2019 at 8:32
  • 1
    $\begingroup$ @Knarpie Well we know the normalized coefficient is t-distributed. The problem is that with more samples you’d expect the standard deviation to decrease and the degrees of freedom to go up...so it’s complicated. $\endgroup$
    – cgreen
    Nov 21, 2019 at 16:40
  • $\begingroup$ It seems, we're look for the same $\endgroup$
    – Christoph
    Nov 22, 2019 at 10:07


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.