# Hypothesis testing: t-test and p-value conflict

I am doing a one-tailed hypothesis test on the exper coefficient. My null is the coefficient is not different from zero and my alternative is that the coefficient is greater than 0. I'm using a 5% level of significance. I get a t-stat of 1.735879 and the t-crit is 1.6759, d.f.=50. So here the t-stat is greater than the t-crit, so it seems I have enough evidence at the 5% level to reject the null. However, the p-value is .0883. Since I am using a 5% level, the p-value suggests that I don't have enough evidence to reject the null.

Could you help me understand this? Have I made a mistake in calculating my t-crit? or does this happen with a low number of observations?

Thank you for looking.

Below is the eviews output.

• Not sure how Eviews works, but this is likely to be reporting a p-value from a two-sided test – sheß Apr 13 '16 at 0:58
• ... and you are using a critical value from a one-sided test, which is almost always a bad idea. – Jeremy Miles Apr 13 '16 at 1:00
• @sheß: Thank you. that is it. Put it up as a solution and I'll mark it for you. – Travis Apr 13 '16 at 1:04
• @Travis If sheß doesn't put up an answer in the next day or so, perhaps you could write one. – Glen_b -Reinstate Monica Apr 13 '16 at 1:34