Statistics: Calculate P-value from T and DF

Dublo7

I need to find an equation that will allow me to calculate a P-value, if I have found a T-Statistic and Degrees of Freedom.

I have found a website that will do it for me, but it doesn't tell me the actual equation, so I can do it myself.

Can someone please help me out?

Thanks.

Gdiguy

Dublo7 wrote: »

I need to find an equation that will allow me to calculate a P-value, if I have found a T-Statistic and Degrees of Freedom.

I have found a website that will do it for me, but it doesn't tell me the actual equation, so I can do it myself.

Can someone please help me out?

Thanks.

There's not an easy to use formula, since it's basically the cumulative distribution function of the t-statistic (which is a pain in the butt)... do you actually have to do it yourself for some reason, or are you just curious?

Dublo7

I'll most likely have to do it during an upcoming test.

I've been told I can use a 'Critical Points of the T-distribution' table to get the P-value, but I can't quite figure out how to use it.

I guess I'll just have to ask my tutor tomorrow.

Gdiguy

Yeah, ok... so the "critical points table" basically just means that you're given a table of t-statistic values that equal a specific p-value (usually 0.1, 0.05, 0.01, etc) for a variety of degrees of freedom. Thus, if you look down the column of degrees of freedom, and you see that the critical value for a p-value of 0.01 is 5, then if your t-statistic is greater than 5 you know that your p-value must be less than 0.01

This works because a lot of the time you don't care what the actual p-value is, you just care if the t-statistic value you have is significant at a specific p-value (which gives you a specific t-statistic critical value / cutoff to look at)

Dublo7

Thanks for the help. I'll give it a try now

Dublo7

So, if my t-statistic ends up being smaller than the critical number, that means it's not statistically significant?

Gdiguy

Dublo7 wrote: »

So, if my t-statistic ends up being smaller than the critical number, that means it's not statistically significant?

Pretty much, yeah... you can't significantly reject the hypothesis that the means are the same

Dublo7

Thanks a lot

musanman

Remember to make sure you're using the right method to test your hypothesis. It could be a left tailed, two tailed, or right tailed tests...which will change how you calculate your critical values.