trendbreaker
trendbreaker copied to clipboard
Analysing epidemic trends using AutoML approaches.
I'm not sure what changed as I've used trendbreaker for months without any issues, but I got this error when I tried to load trendbreaker today (4/13/23). Error: package or...
To apply asmodee to a number of separate timeseries in a grouped dataset so far the dataset first has to be converted to a grouped incidence object. It would be...
Before any release - check #51 is the best approach to parallel.
Both for the basic output, and the one of incidence2 objects. Ideally return a data.frame with: * number of increases in training set * number of decreases in training set...
It feels like this function should move to trendeval, and be exported. Selection is not trivial in the case of `evaluate_resampling` which can use several metrics to assess model performances,...
ASMODEE currently can detect changes when they first occur, but after a while changes become part of the past trend, and are therefore no longer seen as changes. An option...
Ideally, the calibration window should include a single trend which can be modelled by at least one of the candidate models. In practice, this may require selecting a subset of...
## Context Using the range of models below to fit count data at small spatial scales, zero-only time series in the callibration window are common; these trigger an error with...
From Sebastian Funk: RMSE seems a slightly odd choice here - since the accuracy of the method will depend on probabilistic calibration (the outlier detection doesn't care about the mean),...
For `alpha = 0`, the upper and lower bounds should always be `Inf` and `0`, respectively. However: - For `glm`: trendbreaker uses `ciTools::add_pi` internally to find upper and lower bounds....