Is the pvalue from an lmfit table derived from a two-sided or one-sided t-test?

I am using the lmfit to do a linear regression analysis between independent variable x and its relationship with the dependent variable y. The output of this simple model looks like:

y ~ 1 + x1

After using lmfit on the data for x and y variables, a table providing the t-statistic and the associated p-value are given for variable x1.

My question is, does the p-value in this table automatically account for a two-tailed test or does the p-value need to be doubled to account for this? According to the MATHWORKS page, the pvalue for the t-statistic is for a two-sided hypothesis test, so it leads me to believe that the p-value that is reported in the table does not need to be doubled and is already taken into account in the calculation. And if that is true, if you wanted to do a one-sided hypothesis test based on this p-value, you should halve the p-value to get your answer. Could anyone verify these questions/assumptions?

Here is my source on MATHWORKS about lmfit being a two-sided hypothesis test: https://www.mathworks.com/help/stats/fitlm.html#mw_11b75ad7-ce2d-4251-9e8f-2b70dbfe4f4d

Thanks so much!

Answer 1

[Star Strider]

For what it’s worth, when I updated my polyparci function, I used a default two-sided t-distribution to calculate the probabilities, and used fitlm to verify my probability results. It is two-sided.

That is the only sort of verificiation I can provide.

[William]

Thank you Star Strider, that does give me some additional confidence!

The reason I think it is a two-sided test is that one of my output's had a p-value greater than 0.9 (i.e., non-significant), and doubling that value does not make sense anyway (i.e., probability above 1).

Funny enough, I have used the polyfit function on my dataset originally. Perhaps I could try to run it through the polyparci to verify if its two-sided?

[Star Strider]

My pleasure!

I am certain that polyparci will appreciate your downloading it and using it in your research! It allows for one-sided and two-sided tests.

It also produces confidence intervals for the parameters (its original intent, I added the probabilities to it in an upgrade).

[William]

Thank you so much, I just saw your page for that function! Appreciate it and already testing it on my data.

[Star Strider]

As always, my pleasure!

It has a friend polypredci that calculates the confidence intervals on the fit, if you need to do that as well.

Please.let me know if you have any comments or concerns about any of them.