1
$\begingroup$

I have fit a GLM where the response variable represents ratio and used Binomial family. Below is my model

library(foreign)
df <- read.dta("http://fmwww.bc.edu/ec-p/data/wooldridge/401k.dta")
df$prate <- df$prate/100

myglm <- glm(prate ~ mrate + age + sole, data = df, family = binomial)
summary(myglm)

Now I want to get quantile residuals of above fit. As per the discussion in the paper https://www.researchgate.net/publication/2647151_Randomized_Quantile_Residuals, in order to estimate quantile residuals, I need to convert the response variable to standard normal variable.

However, given that in my case I only have ratios (not actual realization of binomial distribution), I could not figure out how exactly I should estimate the quantile residuals in this case.

Is there any direct or approximate method available to obtain such residuals for this type of model fit?

Any pointer will be very helpful.

$\endgroup$
1

1 Answer 1

0
$\begingroup$

You cannot compute randomized quantile residuals without the binomial numbers. In fact you cannot fit a binomial glm corrrectly without them either.

But the dataset you have loaded does include the binomial realizations. You need:

library(foreign)
df <- read.dta("http://fmwww.bc.edu/ec-p/data/wooldridge/401k.dta")
df$prate <- df$totpart / df$totelg
myglm <- glm(prate ~ mrate + age + sole, data = df, family = binomial, weights = totalelg)
library(statmod)
qr <- qresid(myglm)
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

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