This website uses Cookies. Click Accept to agree to our website's cookie use as described in our Privacy Policy. Click Preferences to customize your cookie settings.
Submit your abstract to the call for content for Discovery Summit Americas by April 23. Selected abstracts will be presented at Discovery Summit, Oct. 21- 24.
Discovery is online this week, April 16 and 18. Join us for these exciting interactive sessions.
Created:
Mar 29, 2017 02:52 PM
| Last Modified: Mar 29, 2017 02:55 PM(7534 views)
I am doing an independent sample t-test, and I've encountered a problem with the degrees of freedom. What I'd like to do is use the floor (13 in the attached analysis), not the unrounded (13.65148), df, in calculating the confidence interval and p value. Is there a setting I can change in JMP so that it uses the floor of the df? Thank you.
You have two choices. First of all, you could use the built-in function to compute your t test. Use the t quantile( p, df ) function in a column formula or in a script. Truncating the DF is more conservative but it seems arbitrary to me. Second of all, you can simply use a different alpha. Click the red triangle next to Oneway and select a new significance level for computing the confidence intervals. Of course, you can apply a new alpha value in your decision based on the given p-value without change. This way seems better to me because you know how conservative the new t test is.
I believe that the fractional degrees of freedom for the error arises when you use the unpooled version of the t test. This is correct for that case. The pooled t test does not adjust the DF for the error and is correct if the assumption of equal variance is valid.
Thank you for your reply. I should clarify, though, that I am assuming unequal variance between the two populations, so I do not think the pooled variance is correct here. In order to be more conservative, I would like JMP to use the floor of the df in the calculations for an unpooled t-test. Is this possible?
You have two choices. First of all, you could use the built-in function to compute your t test. Use the t quantile( p, df ) function in a column formula or in a script. Truncating the DF is more conservative but it seems arbitrary to me. Second of all, you can simply use a different alpha. Click the red triangle next to Oneway and select a new significance level for computing the confidence intervals. Of course, you can apply a new alpha value in your decision based on the given p-value without change. This way seems better to me because you know how conservative the new t test is.
Thank you. I've made a table that gives the p and confidence interval from given values. My professor for this class prefers rounding down the df, as it is an introductory course and we do many calculations by hand and then look up the value in a table with only integers for df.
