EDIT: Sorry, this post was rambly before - this is my first question and the instructions said to explain exactly what you were doing. I hope it is clear what I was asking now. And I hope it is appropriate to ask for this help (just need a one word answer really).

QUESTION: I have 87 degrees of freedom. I have to look up the critical value in an F-table. It has 60 and 100 for the second degree of freedom. Should I round up, or down, or interpolate?

  • $\begingroup$ There is a smell of homework. $\endgroup$ – Yves Daoust Dec 21 '17 at 9:28
  • $\begingroup$ Fair enough, sorry. I will email my lecturer - I just didn't want to bother her over Christmas with another question! She is very nice though :) $\endgroup$ – Rachel Kirkland Dec 21 '17 at 9:30
  • $\begingroup$ People here will be eager to answer narrowed down questions, showing more precisely where you are stuck. But they won't like to do all the work for you. $\endgroup$ – Yves Daoust Dec 21 '17 at 9:35
  • $\begingroup$ What is the first DF? // Tradition is to round down when using printed tables, because that is 'conservative' (less likely to lead to unwarranted rejection). Occasionally, I have seen the use of 'reciprocal interpolation' (take reciprocals of the two tabled values, interpolate, take reciprocal), but can't be sure how well it would work in your case. // Nowadays, people often use software. For example, if ndf = 4 and ddf = 70, then in R statistical software qf(.95, 4, 70) returns $2.502656$ as the value that cuts 5% from the upper tail of F(4, 70). // Software also gives lower-tail values. $\endgroup$ – BruceET Dec 22 '17 at 3:53
  • $\begingroup$ ... Lower-tail values are required for confidence intervals. There is a method of for getting them from printed tables (reversing DF's and tails, and taking reciprocals). It may be mentioned in your text. In R, 1/qf(.05, 70, 4) returns $2.502656$. // Yes, this does have the 'smell' of homework, but my guess is this question would be about the very last step in a longer problem, not the essence of problem itself. // Finally, I'll mention that some statistical calculators give reasonable approximations of quantiles of F. $\endgroup$ – BruceET Dec 22 '17 at 4:04

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.