ABSTRACT
I will use the t-test to illustrate the points but the principle will be true for other tests, including non-parametric tests where probabilities are shown for given sample sizes rather than df.
When the tables do not have probabilities for the exact degrees of freedom for the test you have conducted then a quick initial check of the statistical significance is to note whether the t-value for the next lowest df is statistically significant. If it is then it will also be significant with the correct df. For example, if df = 45 then a t-value of 1.7 would be statistically significant at α = 0.05 for a one-tailed test because the critical t-value for df = 40 is 1.684. On the other hand, if the t-value is not significant with the next highest df then it will not be for the exact df. Accordingly, if df = 45 and the t-value was 1.6, then it would not be statistically significant at α = 0.05 for a one-tailed test because the critical t-value for df = 50 is 1.676. If using these approximate methods does not tell you whether the result is statistically significant, then use linear interpolation.
