Calibration Plot

[Pippo Bullo]

I am attempting to perform a calibration plot for a nomogram based on a logistic regression.

Unfortunately I am not able to found on the manual the code to build it. The outcome should looked something like that:


Untitled.tif
Do you know the code for that? thanks

[Rich Goldstein]

your graph is unreadable (by me at any rate) and I do not look at binary attachments - please see the FAQ on how to show graphs

on the main question, I note that there are several ways to do "calibration plots" and many have been discussed on Statalist - please search the archives;

here is one simple way:

estimate model
predict newvar
lowess depvar newvar, addplot(function y=x, range(0 1)) legend(off)

note that the above has recently been studied in "Austin, PC and Steyerberg,
EW (2014), "Graphical assessment of internal and external calibration of
logistic regression by using loess smoothers," _Statistics in Medicine_,
33: 517-535

[Pippo Bullo]

Thank you for your answer. Can you see the graph now? Is it the same outcome you suggested with your code?

[Marcos Almeida]

Hello Pippo,

Welcome to the Stata Forum!

I suggest you take a look at the user-written - nomolog - and, if you wish further examples, here you have a link to the Stata Journal:http://www.stata-journal.com/sjmatch...eyword=nomolog

Recently, I used - nomolog - for a logistic model and it worked gracefully!

Best,

Marcos

[Pippo Bullo]

Thank you Marcos! I used the guide from the same author you suggested me to built my nomogram: http://www.zlotnik.net/stata/nomograms/
Unfortunately there was nothing about the calibration plot that I think it should included in this type of model

[Marcos Almeida]

Great, Pippo, but I wonder whether you couldn't use ROC curves for the plot. If not, this link (https://www.sealedenvelope.com/stata/hl/) with the user-written hl may perhaps fulfill your needs.

Best,

Marcos

[Pippo Bullo]

Unfortunately hl seems different from the other type of calibration plot used in my area (I tried to do it following this: https://www.sealedenvelope.com/stata/hl/) .

This is for the purposes of my research the best look:

[Rich Goldstein]

sorry, I don't know what data you used for #3 above and therefor can't answer your question (but maybe I misunderstand what your question is?)