topic Re: Model comparisons in Discussions

Model comparisons

tox — Fri, 09 Jun 2023 00:33:35 GMT

I have dose response data for two different treatment groups (see table below), and want to compare the dose response curves. I have run the analysis in the screenshot below, and the effects test shows that both dose and the treatment (Tx) are significant factors. What statistical test is JMP using to do the comparison shown under Effects Tests? Any help would be greatly appreciated.

Re: Model comparisons

SDF1 — Fri, 14 May 2021 14:43:15 GMT

Hi @tox ,

You can read more about Effects Tests here. I think in your case it's basically performing a t-test on the estimate for that factor and comparing that value to the null hypothesis, which would be the estimate for the factor is 0. So for your case, the p-values are small and you can reject the null hypothesis and conclude there is enough evidence to accept the alternative hypothesis, which is the factors and their estimates are significant in the model.

Hope this helps,

Re: Model comparisons

Mark_Bailey — Mon, 17 May 2021 16:56:31 GMT

It is not a t-test.

The column header in the Effect Tests is "L-R ChiSquare." This header indicates that it is a 'likelihood ratio test." The test statistic is twice the change in the likelihood function when the term is added to the model. This statistic has a chi square distribution. It is analogous to using a t-test of the estimate versus zero or a F-test of the sum of squares, which cannot be used in this case because the response is not continuous. It is categorical (counts), so it must be based on probability.

Re: Model comparisons

tox — Mon, 24 May 2021 15:17:50 GMT

Thanks for the answer on this. One other quick question - is there a way to have JMP display the prediction expression for the curves? I know this is a menu option under other models, but I don't see it here. Thanks again!