1
$\begingroup$

Now, I'm not going to bother you any longer, this is my last question (and thanks so much for all of your help): Is there a way of calculating an ordinal logistic regression "by hand" without going insane? Maybe it's just me but it seems incredibly difficult.

I'm not opposed to shortcuts in R, I just don't want to use the simple ordinal regression command for once.

I'm sorry if this question isn't very specific, I'm basically clueless.

$\endgroup$
5
  • $\begingroup$ Yes, you can do it using pens and papers only, as long as you get enough time $\endgroup$ – Deep North Jul 1 '17 at 21:19
  • $\begingroup$ With pen and paper? You'd go nuts with any nontrivial sample. It's not clear what counts as acceptable "shortcuts in R" in your case, if you don't want to use preexisting ordinal-regression routines. What are you actually trying to accomplish? $\endgroup$ – Kodiologist Jul 1 '17 at 22:51
  • $\begingroup$ It's for practice, I was thinking about calculating one example by hand (in addition to the usual command) to get a better understanding. I did the same thing with kendalls tau (correlation) and just used R to find the number of concordant and discordant pairs. But I'm not going to go through my dataset of over 900 observations to count every single (possible) outcome. If that's what I'll end up having to do. $\endgroup$ – Amb Jul 2 '17 at 10:58
  • 1
    $\begingroup$ @Amb To "get a better understanding" you will be much better off writing your own code for fitting some simple version of the model than "doing it with pen and paper". $\endgroup$ – Jarle Tufto Jul 2 '17 at 14:14
  • $\begingroup$ Ok, so it doesn't really make sense to calculate the regression by hand. Thanks so much for your help! $\endgroup$ – Amb Jul 2 '17 at 14:38
1
$\begingroup$

The ordinal package is probably the most common for fitting ordinal regression in R. You can get some sense of how it fits models by reading the document linked below (first link), and by the other support documents on CRAN (second link below). Specifically, the package uses cumulative link models fit with maximum likelihood estimation. This is method is probably not something you want to try to reproduce by hand, but the documents might make the procedure comprehensible.

https://cran.r-project.org/web/packages/ordinal/vignettes/clm_intro.pdf

https://cran.r-project.org/web/packages/ordinal/index.html

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.