ArboMAP 2.3

Data are now associated with a "believable" parameter, which requires the user to tell ArboMAP which data can be trusted when. This is less clever than trying to infer the limits of credibility from the files, but it is also more robust against, for example, data missing in the middle of a time series rather than on its edges.

We no longer attempt to figure out when the user wants to do predictions - the user chooses the list of years for which estimates should be produced.

Shapefiles are now downloaded with the tigris package during the run, rather than requiring the user to produce and store them before running the program.

Estimates of random effects in the mosquito model are now more robust. Occasionally, numerics would be forcibly converted to factors whenever the numerics were all integers, for example, and this would cause problems in the estimates. We now allow missing mosquito data in the middle of the time series.

We have now adjusted the model to include indicators whenever a year has believable human data that is to be modeled, but there is no associated mosquito data. In this case, the missing mosquito data are assumed to be average and the intercept is simply fit with a constant for that year.

There have been changes in language throughout to better explain the workings of the program.