https://community.jmp.com/t5/Discussions/Choice-model-platform-and-marginal-effects/m-p/522766#M74751 <P>Hi everyone,</P><P>I have used the choice model platform to design and analyse choice experiments. However, at the analysis step, there are some parts of JMP platform output that are not clear, and I wonder if you can provide an answer:</P><UL><LI><SPAN>It is </SPAN><STRONG>not clear how marginal effects are calculated in JMP</STRONG><SPAN>.</SPAN></LI><LI><P><SPAN>More specifically, how do we obtain marginal probabilities from marginal utilities since neither the utility function nor the probability distribution is ever specified in JMP?</SPAN></P></LI><LI><SPAN>What is the </SPAN><STRONG>difference between marginal effects and predictive probabilities</STRONG><SPAN> (that we </SPAN><I><SPAN>should </SPAN></I><SPAN>get when using “</SPAN><STRONG>Probability Profiler</STRONG><SPAN>”) in this model?</SPAN></LI><LI><SPAN>There are </SPAN><STRONG>no indicators of statistical significance for marginal effects nor predicted probabilities</STRONG><SPAN>.</SPAN></LI></UL><OL><LI><P><SPAN>How can we test that results from one model (e.g., a baseline model with just attributes) are significantly different from another (e.g., a model with an individual level covariate)?</SPAN></P><BR />I reckon these are rather specific questions, but if someone from the JMP team could answer, it would greatly help avoid a black-box use of the software.</LI></OL><P>&nbsp;Thank you</P><P>&nbsp;</P><P>&nbsp;</P> Fri, 09 Jun 2023 00:52:27 GMT gav2013 2023-06-09T00:52:27Z https://community.jmp.com/t5/Discussions/Choice-model-platform-and-marginal-effects/m-p/522766#M74751 <P>Hi everyone,</P><P>I have used the choice model platform to design and analyse choice experiments. However, at the analysis step, there are some parts of JMP platform output that are not clear, and I wonder if you can provide an answer:</P><UL><LI><SPAN>It is </SPAN><STRONG>not clear how marginal effects are calculated in JMP</STRONG><SPAN>.</SPAN></LI><LI><P><SPAN>More specifically, how do we obtain marginal probabilities from marginal utilities since neither the utility function nor the probability distribution is ever specified in JMP?</SPAN></P></LI><LI><SPAN>What is the </SPAN><STRONG>difference between marginal effects and predictive probabilities</STRONG><SPAN> (that we </SPAN><I><SPAN>should </SPAN></I><SPAN>get when using “</SPAN><STRONG>Probability Profiler</STRONG><SPAN>”) in this model?</SPAN></LI><LI><SPAN>There are </SPAN><STRONG>no indicators of statistical significance for marginal effects nor predicted probabilities</STRONG><SPAN>.</SPAN></LI></UL><OL><LI><P><SPAN>How can we test that results from one model (e.g., a baseline model with just attributes) are significantly different from another (e.g., a model with an individual level covariate)?</SPAN></P><BR />I reckon these are rather specific questions, but if someone from the JMP team could answer, it would greatly help avoid a black-box use of the software.</LI></OL><P>&nbsp;Thank you</P><P>&nbsp;</P><P>&nbsp;</P> Fri, 09 Jun 2023 00:52:27 GMT https://community.jmp.com/t5/Discussions/Choice-model-platform-and-marginal-effects/m-p/522766#M74751 gav2013 2023-06-09T00:52:27Z