1
$\begingroup$

If I have hypothesized that a variable $x$ will have a positive effect on $y$, but the analysis shows that the effect is negative, but not significant, how should I interpret these results? Should I say that the results do not support my hypothesis, yet that it is not conclusive whether this is true only for the sample or also for the entire population?

$\endgroup$
  • $\begingroup$ It also depends what your standard errors are, how much power your statistical tests have. If your tests have no power, failing to reject the null doesn't mean much of anything. On the other hand, if your standard errors are minuscule and you fail to reject the null, it suggests the effect is near zero. $\endgroup$ – Matthew Gunn Aug 31 '16 at 1:30
4
$\begingroup$

Generally, if a coefficient is statistically insignificant, then we say the estimate is too imprecise to make any claim about suggestive correlation.

However you should do a hypothesis test on the effect of x being positive, not just zero (which is likely what your regression output gave you). This may change your inference.

So $H_0=\beta>0$, with $H_a\leq 0$

$\endgroup$

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.