Discrepancy between implementation and paper?

[torfjelde]

I'm currently trying to reproduce the model in another PPL and I'm slightly confused when comparing the model and code.

In the paper (p. 20) you can find the following: while in the code we have https://github.com/ImperialCollegeLondon/covid19model/blob/5c756a69f9d73e28ffe37a15b6cccc157dcf5160/stan-models/base.stan#L67-L68 https://github.com/ImperialCollegeLondon/covid19model/blob/5c756a69f9d73e28ffe37a15b6cccc157dcf5160/stan-models/base.stan#L34 which means the code is implementing Exponential(0.03) and Exponential(1 / τ) rather than Exponential(0.03) and Exponential(τ) as specified in the paper (or with inverse parameters, depending on which parameterisation you're using in the paper).

So I'm wondering if this is a mistake in the code or just a typo in the paper?:)

Also, from the above snippet from the paper, it seems to me that y[m] should be a random vector of length num_impute or N0 (6 in paper) for each m, while in the code all the imputed values seem to be the same value: https://github.com/ImperialCollegeLondon/covid19model/blob/5c756a69f9d73e28ffe37a15b6cccc157dcf5160/stan-models/base.stan#L39 Is this correct?

Thanks!

(Btw, really appreciate you open-sourcing this!)

[s-mishra]

Hi @torfjelde, thanks for pointing this out. There is a typo in the paper. It should have been exponential(1/tau). Also yes all imputed values are same for all N0 day but it is specific to a country by virtue of y being a vector of length M (where M is the total number of countries)

[torfjelde]

Aight :+1: Thanks for the reply!:)