topic Re: Degrees of freedom discrepency in Discussions

Degrees of freedom discrepency

rcspencer — Mon, 27 Mar 2017 20:51:07 GMT

I am running a mixed model effect on two different dependent variables. Both variables cosist of continuous, real numbers with no missing data. However, I get two different degrees of freedom depending on which dependent variable I use in my analysis. Since they both contain the same number of responses, I don't understand how this is happening. Any insight? IIt also looks like there are only 5 responses shown plotted for the dependent variable Prop, while d' displays all 6 data points. Attached is a subsset of my data which displays the same errors.

Re: Degrees of freedom discrepency

rcspencer — Mon, 27 Mar 2017 21:17:29 GMT

I should specify that it is actually the DFDen that is changing.

Re: Degrees of freedom discrepency

Mark_Bailey — Tue, 28 Mar 2017 14:22:10 GMT

I assume that you are using the default REML procedure and its default options to fit this model. As such, the degrees of freedom for the denominator are adjusted and the adjustment depends on the response variable and parameter estimates.

Re: Degrees of freedom discrepency

Mark_Bailey — Tue, 28 Mar 2017 14:26:03 GMT

Select Help > Books > Fitting Linear Models. Search on REML to see background information about the estimation merthod and the reports. Search on Kackar-Harville to see details about this adjustment.

Re: Degrees of freedom discrepency

rcspencer — Wed, 29 Mar 2017 03:21:44 GMT

Thanks so much!