1
$\begingroup$

The title pretty much says it. Sorry, it's so unspecific, I am not asking the question for myself and I am not familiar with stats.

$\endgroup$
7
$\begingroup$

p value is a function of effect size and sample size.

A larger effect size and a larger sample size will yield a lower p value (the question seems to imply the opposite).

$\endgroup$
  • 1
    $\begingroup$ Thanks for the answer! Yes, the question actually what reduces the p-value, I wrote the wrong thing. I changed it now. $\endgroup$ – ChriBa Jan 10 '16 at 14:20
4
$\begingroup$

Maybe that is already subsumed in Peter Flom's definition of the effect size, but a higher signal (for me, that is the effect size) to noise ratio: for example, in a regression context, your t-statistic would be $$t=\frac{\hat\beta_j-\beta_{j0}}{s.e.(\hat\beta_j)}$$ where $s.e.(\hat\beta_j)$ is the $j$th diagonal element of $s^2(X'X)^{-1}$. Here, $s^2$ is an estimator of $\sigma^2$, the error variance. A larger $\sigma^2$ will, all else equal, tend to produce larger $s^2$ and thus translate into a smaller t-statistic and hence a larger $p$-value.

$\endgroup$
1
$\begingroup$

The p-value is defined as the probability of obtaining results at least as extreme as your actual data under the null hypothesis. The same data will yield higher or lower p-values depending on what exactly the null is, eg one or two sided, multi- vs univariate, etc. For example, in a multiple regression, an omnibus test may come out significant even though the slopes individually do not.

$\endgroup$

Your Answer

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

Not the answer you're looking for? Browse other questions tagged or ask your own question.