0

I have recently submitted a paper that demonstrates that a categorical outcome variable (0 = irregular verb production; 1= regular verb production) is significantly predicted by a continuous variable (phonological similarity to existing verbs). Specifically, a glmer model showed that production of regular verbs was significantly more likely when the similarity measure increased.

I plotted this relationship by plotting MEAN production on the Y axis, and the similarity measure on the x axis. However, a reviewer has asked, rather vaguely, that I "plot the axis in a log-odds scale". I'm struggling to work out exactly what is meant here, and am wondering if this is a standard requirement for plotting the relationship between a categorical outcome variable and a continuous variable?

Is it as simple as log transforming both x and y axes, then doing the same plot (i.e., x axis = natural log transformed mean production; y axis = natural log transformed similarity measure)? If I do this, the plot looks identical to the non transformed version, so I wonder if I should try something else?

Many thanks for any help. Ryan

Ryan
  • 1
  • 2
    if `p` is the probability of the outcome, then the log odds is `log(p/(1-p))` or `logit(p)`. SO is not the site for such questions, but rather cross validated. Here programmatic problems and solutions are discussed. If you would like post an example of the data, what you tried and what you are after. – missuse Oct 01 '17 at 19:17
  • If you want this to be an SO question and get helpful answers, then you should at least post some code you have tried so far and data for people to work with. As @missuse pointed out, SO is for programming-specific problems. – acylam Oct 01 '17 at 20:01

0 Answers0