p-value

[travelin.man119]

If the p-value of a coefficient is 0 it implies a 100 percent chance it is significantly diff from 0 and a p value of 1 implies it is not sig, diff from 0, correct?
Also can the p-value be compared directly to the t-stat to tell us anything. The p-value represents the lowest value we can still reject the null so if the t-stat is lower than p does that tell us we can reject the null as well?

 

[tickersu]

travelin.man119 wrote:
If the p-value of a coefficient is 0 it implies a 100 percent chance it is significantly diff from 0 and a p value of 1 implies it is not sig, diff from 0, correct? No, p-values you see are rounded (so it isn’t really 0). You can only conlclude that the p-value is much smaller than any conventional level of alpha, so it is very likely that the true value of the coefficient is different from zero. However, this is not a 100% guarantee.
Also can the p-value be compared directly to the t-stat to tell us anything. The p-value represents the lowest value we can still reject the null so if the t-stat is lower than p does that tell us we can reject the null as well? You can’t compare the p-value directly to a test statistic of any kind. The p-value represents the probability of seeing a result more contradictory to the null hypothesis given that the null is true (observed significance). So, the p-value is the smallest level of alpha (chosen significance) at which you can still reject the null hypothesis.
Compare apples to apples. Probabilities to probabilities (alpha/chosen significance threshold to an observed p-value) and standardized values to standardized values (test statistics to critical/cutoff values).

 

[ahungerford]

Don’t try to compare the p-value to the t-statistic directly.
The most important things to remember about the p-value are:

• if the p-value < alpha (significance level), you reject the null hypothesis (reject H0)
• the smaller the p-value, the stronger the evidence AGAINST H0.