-
Notifications
You must be signed in to change notification settings - Fork 13
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
estimate growth rate r using Poisson glm instead of log-linear model #125
Comments
similar issue, different package: |
Yes, definitely agreed! This is what I am currently using - quasi-Poisson actually. I was not aware of EpiNow .. there is some duplication there, and I am not sure how to handle this going forward. |
yes, if there is overdispersion quasipoisson or negative binomial regression would be best. |
Related issue I just posted: epiforecasts/EpiNow#75 |
The way the exponential growth rate is estimated here currently (in
fit()
) is to use a log-linear model (linear model similar tolm(log(incidence) ~ date)
. For the log to work you need to remove the 0s or replace them with a small positive value.A better way to estimate the growth rate would be to use a Poisson glm of the incident cases (
glm(incidence ~ date, family = poisson)
or similar). In addition to the more appropriate error model for count data this can handle 0s in the data natively.Here are some references:
https://besjournals.onlinelibrary.wiley.com/doi/10.1111/j.2041-210X.2010.00021.x
https://bmcmedinformdecismak.biomedcentral.com/articles/10.1186/1472-6947-12-147
And here is how this is done in the R0 package, where you can chose between poisson and log-linear models, with a default to poisson (https://github.com/cran/R0/blob/master/R/est.R0.EG.R):
The text was updated successfully, but these errors were encountered: