# R: Interpreting mlogit coefficients

Edit: The subset of the dataframe I provided way giving a different error, so I've replaced the pastebin entry with the full data frame.

Here's the top of the data frame, then read into mlogit format by clogitdf <- mlogit.data(df1, choice= "y", shape="long", alt.var="id_var")

http://pastebin.com/ytisEDfD

Variables:

PriceNorm = Price of brand normalized to size (eg. $/oz.) Inc = Income y = (Consumer's choice: Tide, Surf, or Wisk) Here's the model: reg2 <- mlogit(y ~ PriceNorm | Inc, data=clogitdf, id.var="id_var", reflevel="surf") The goal is to model how consumers prefer the three brands Tide, Surf, and Wisk. When I run the model, I get a positive coefficient on PriceNorm = 13.8. I'm confused because my interpretation of that coefficient is that for a$1/oz. increase in the price of Surf, consumer utility increases by 13.8.

It would help me if you can check my work and make sure the model is doing what I intend it to, so I know whether to trust the PriceNorm coefficient or not.

Edit: I spoke to my professor today, and PriceNorm should definitely be negative. I think I need to include the dummy variables, but when I do, I get an error saying the model is exactly singular.

• Are you sure that it's not PriceNorm + Inc? – Dimitriy V. Masterov Jan 30 '15 at 22:44
• @DimitriyV.Masterov Inc doesn't change with the question, only the survey respondent, so I think that makes it an individual specific variable which comes after the first | – Mako212 Jan 30 '15 at 23:24