As I am not able to regularly update the content here, and this blog is valuable to me, I have decided to give it a little more value by transforming it into my personal scientific monitor.
I give myself a command to publish a new entry here at least twice a week in order to see what I have learned, what I have discovered, and how the things are going on in general.
Now I am in Germany, realizing my doctoral practice in Munich. At work I deal with the proportional odds statistical model and make a lot of use of the R software and of its VGAM package in particular. Everything is new to me, due to the fact I am working with Mac since recently. This is my first Mac, and to me it's freaky! I hope I'll get over it soon, and we'll make friends.
Everybody who'll ever use VGAM for proportional odds model prease bear in mind two things:
- ...family=propodds doesn't really work; use the family=cumulative(parallel=TRUE) instead;
- while interpreting the model's summary remember that VGAM accords to the following approach:
logit[P(Y<=y|x)]=a - bx
so the coefficient's impact should be consumed with the opposite sign.
VGAM is tricky but it is a really good stuff.
