Sam Abbott
Sam Abbott
It is desirable that random walks are supported with multiple lags and orders. This issue is a placeholder for a full design sketch. See `rw` and `construct_rw` for the relevant...
As suggested by @jbracher a normal observation model may approximation the negative binomial and Poisson observation models well when counts are high. A more generic normal observation model may also...
As suggested by @jbracher a gamma Poisson mixture may be closer to the commonly used model structure of a negative binomial observation model for the final observed counts and a...
## Feature request Splines are often useful for fitting non-linear relationships and can be more efficient than AR processes etc. It would be nice if `epinowcast` was able to support...
Our model is now quite complex and only likely to get more so. It might be sensible to think about ways of splitting up our model vignette further (as done...
From @parksw3 - I think there needs to be some more clarification between the relationship between $N$, $\lambda$, and $n$ before you get to the default model. In particular, the...
From @parksw3: `germany-age-stratified-nowcasting.Rmd` - "reporting delays are fixed across age groups and time" Do you mean that the delay "distribution" is fixed? - I can't figure out where you're specifying...
As we outlined in our style guide (`STYLE_GUIDE.md`) `epinowcast`, - We support inputs that are coercible to `data.table` objects using `data.table::as.data.table()`. This includes `data.frame` and `tibble` objects. This should be...
From our `STYLE_GUIDE.md`: - For external functions we aim for the output to be a `data.table` object if possible unless a custom class is used (which we generally aim to...
In #165 @sbfnk used posterior predictions as data to check the model. This is generally considered good practice (as is using prior predictions (which we can achieve by skipping the...