How to Display 10% Significance in a Conjoint Experiment Graph?

[Bruna Ribeiro]

Hello everyone,

Thank you in advance to anyone who can help.

I am writing a paper in which I used a conjoint experiment as my methodology. To generate the traditional conjoint experiment graph, I ran the following Stata command:

conjoint choice candidate_gender candidate_age candidate_educ candidate_maritalstatus candidate_children candidate_experience candidate_localties candidate_ideology, est(amce) id(ID) graph(0)

The graph I get is exactly what I need, but it only displays significance levels below 10% (p < 0.10). I would like it to also include significance at the 10% level (p = 0.10).

Does anyone know how I can modify the command to include this information in the graph?

I appreciate any help!

[Andrew Musau]

conjoint is from the Stata Journal (FAQ Advice #12).This is a command that I do not use, so I cannot directly answer your question. However, if the coefficient has a p-value of 0.101, it is technically not significant at the 10% level. Claiming that it is significant would be stretching the truth.

For some Stata commands, there is a -level()- option that allows you to specify the significance level. For example, -level(89.99)- would correspond to a 10.01% significance level, whereas -level(90)- might be the default, corresponding to a 10% level. If this option does not exist and no one else responds, you may need to reconsider your interpretation or contact the program authors for further advice.

[Hemanshu Kumar]

The graph I get is exactly what I need, but it only displays significance levels below 10% (p < 0.10). I would like it to also include significance at the 10% level (p = 0.10).

I have to say I don't quite understand your question. If you are worried strictly about p = 0.10, remember that it is a zero probability event that you will have a p-value of exactly 0.10.

I too have never used conjoint, so I'm not clear whether what follows is the solution for you. In particular, I don't know if your problem is with only selected coefficients being shown, or with confidence intervals of a specific level. If the latter, then this may help:

I notice that conjoint in turn calls coefplot to make its graphs, and as the help file for conjoint states, you can even show the graph command by following the call to conjoint with

display _asis "`e(graph_code)'"

Further, as the help for coefplot shows, it is very much one of the commands that Andrew referred to in #2: it has an option levels() to specify the level of the confidence intervals. So you can tinker with that as needed.

Does that solve your problem?