Choice analysis output

[#1]

Hello,

I am using jmp for discrete choice analysis and am confused about the output. For the parameter estimates, they appear to be comparing different attribute levels within the parameter; what is jmp doing here?

As well, when I incorporate subject factors, only 2 of the 3 subject factors are presented on the output, why is one of them being ommitted?

Lastly, what does the logworth values represent? Can these values be used to appreciate the relative significance of each attribute on respondent decision making?

[#2]

The Likelihood Ratio Tests report is useful when you want to decide if a term in the model is significant. The Parameter Estimates report is useful when you want to know if an estimate is significant.

Categorical predictors / factors appear in the linear model in a way such that the estimates across all the levels must sum to zero ("effect parameterization"). Therefore, the estimate of the parameter for the last level is equal to the negative of the sum of the other estimates. You can click the red triangle at the top and select Estimates > Expanded Estimates. (I think)

Logworth is the negative log (base 10) of the p-value. (It is the p-function of the p-value, like pH is the p-function of the Hydronium activity.) Because p-values range between zero and one, sometimes they differ by many orders of magnitude. Some users find logworth easier to gauge the significance of estimates in such cases. A change of 1 logworth is a change of one order of magnitude. Note that such users often consider logworth = 2 (p-value = 0.01) a reasonable reference for comparison but you should always determine the level of significance that is meaningful to you.

[#3]

Hi,

I'm new to JMP and statistics, and I'm doing a similar research using Choice Design platform and have some questions pertaining to this thread so I post them here.

I saw in this case study (  ) that the choice model being estimated in choice design platform is a multinomial logit model. From what I've read, for MNL we would want to know odds ratios instead of coefficients (or parameter estimates as called in JMP) of each attribute level to compare their effects. Does JMP do this for you? If so, where can I find it?