F (18, 95) = 6.009, p < 0.0005

I was looking into a research paper, and this F value was provided when using a regression model. I do not understand, what is the need to discuss F value in a paper? It is just an intermediary value. Or, does it really help in understanding some fact other than observed from correlation/regression results.


Ambassador to the humans
From an applied perspective it's not really important. I typically don't even process the test statistics mentally when they are presented like that. With that said think presenting them is important. It makes the analysis at least a touch more transparent. If I somehow got access to the data one could check that the test statistics presented match what you could get from analyzing the data yourself. Sometimes it also helps give an idea about the relative size of the p value and in doing so can help give an indication if something smells fishy.


Less is more. Stay pure. Stay poor.
I duplicate Dason's statements and say that some more social sciences have a greater tradition of reporting such measures. Which I think is just an artifact.

Some authors may feel that they are presenting more results that look technical. Though, what this showed me was hey you have quiet a few predicts given your sample size. I have no idea the context around this model, but it could be overfitted.


Super Moderator
Reporting F statistics that way is dictated by APA style. But I think it can be useful to report them for a couple of reasons:
1) If you haven't reported the actual means/effect sizes, the F statistic can be more useful as a way to back-calculate the effect size than the p value (since the p value will often be a very small number that may not reported precisely, e.g. "p < .001")
2) It allows readers and reviewers to check at least a small part of your calculation process. People report p values that don't line up with the reported intermediary test statistics a surprisingly large proportion of the time (anecdotally)


Super Moderator
F (18, 95) = 6.009, p < 0.0005

It suggests to me - at face value - that the sample size is too small (N=114) with 18 predictors for an OLS regression model.