Hello. I am new here. I have a 2^3 full factorial screening DOE with 3 centre-points. I am fitting a second-degree interactions model, so there are 4 df left for the Total Error; 2 for lack of fit, 2 for pure error. When I display the Parameter Estimates, the Std Error using the "Default" Error Specification is correct and the 95% CIs use a t-value of 2.776 based on 4 df. When I switch to "Pure Error" Error Specification, the Std Error is correctly updated but the 95% CIs still use a t-value of 2.776 based on 4 df. I would have thought the CIs would have used t-value of 4.303 based on 2 df from the 3 centre-points. When I try to use "Specified" Error Specification and enter the center-point error variance with 2 df, the 95% CIs still use a t-value of 2.776. In fact, no matter what I specify, the Std Error is updated accordingly, but the CIs are always based on 4 df. Is this a mistake in JMP, or am I missing something? I am using 17.1.0. Data table is attached. Thank you.
