- function (that calls
predict_delta_comps()
) that does the minutes in the day specifically by default (also deltas in minutes rather than proportions in output) - parameters in the model output should state what the ILRs are (what activities are the numerator/denominator of the ratios)
- additional commenting
- fixed error in
append_ilr_coords()
that would create errors when a single row data.frame was passed (failing unit tests and examples). Seemed to be a new way thecompositions
package handled data potentially, or the move to R 4.0
- the main function
get_plus_minus_changes()
has been more sensibly renamed topredict_delta_comps()
- refactor and modularisation of
predict_delta_comps()
- the two types of reallocation (
prop-realloc
andone-v-one
) are working correctly, unit tested - includes more checks to throw errors for obvious malfunctions
- the mean composition now correctly uses the geometric mean (on the simplex) and not the naive arithmetic mean of the compositions
- added plotting function
plot_delta_comp()
(see '5. Output and plotting results' below) predict_delta_comps()
now removes rows withNA
values in input datasets (and warns the user)- the mean composition and the resulting predicted outcome with confidence interval is now produced by
predict_delta_comps()
(see '5.1 Prediction for the mean composition' below) predict_delta_comps()
now checks for compositional components of 0 [or non-sensical negative values] (geometric mean incompatable)- completed general testing and unit tests