0
      Source |       SS       df       MS              Number of obs =     300
-------------+------------------------------           F(  5,   294) =16580.50
       Model |  1.1499e+10     5  2.2998e+09           Prob > F      =  0.0000
    Residual |  40778600.2   294  138702.722           R-squared     =  0.9965
-------------+------------------------------           Adj R-squared =  0.9964
       Total |  1.1540e+10   299  38593905.2           Root MSE      =  372.43

Should I be worried about this- does your F statistic size matter? How do i find out what the p value of my model is - is it the p value attached to _cons in the bottom of my coefficients?

what do i test this F stat against and what warrants significance?

Heather
  • 1
  • 1
  • The significance is determined by the p-value, which also appears in the output. If that isn't perfectly clear, then please consult our [threads on p-values](https://stats.stackexchange.com/questions/tagged/p-value?sort=votes&pageSize=50). Applications to regression and the F-test can be found at https://stats.stackexchange.com/search?q=%5Bp-value%5D+F+test+regression. – whuber Apr 12 '18 at 18:22
  • 1
    A result can be too good to be true. Does your science tend to produce near perfect explanations (e.g. highly controlled experiments with minimal measurement error)? If the data are economic, $R^2$ that big is (unfortunately) likely to be spurious, but whether the story is tautology, an outlier, or something else is difficult to say. Plotting a scatter plot matrix of the raw data or observed versus predicted or residual versus predicted are among the possibilities. All of these are easy in Stata or any other good software, but ask on Statalist for how if you need to. – Nick Cox Apr 12 '18 at 18:41
  • testing at the 1% significance level, the regression and each coefficient is significant. my question regarding the F statistic - i mean does it matter that the F statistic is so large – Heather Apr 12 '18 at 19:42
  • Yes, it matters. The figures of merit are linked. I focus on $R^2$ because it has a concrete interpretation, but $F$ is enormous and the $p$-value minute. A fit that good is too large to be plausible as an original statistical finding without extra evidence. – Nick Cox Apr 13 '18 at 00:18

0 Answers0