<NeedARTDec31 MV>, and so it was fine to extract the total
value. Now child ART is recorded, and so need to sum the adult age groups only.ylim in plot_prv_pos_yld(). This addresses issue with this plot not showing in interface and preventing plot downloads when data NA values are simulated.vroom::vroom() to read PJNZ files; much faster for reading .DP file.read_specdp_demog_param(), ensure no zero totals when normalising age-specific fertility
distribution and net-migration age distribution.Updates for 2023 UNAIDS estimates:
Other:
data.table dependency.Updates for Spectrum transition from mid-year projection period to calendar year projection period. Implemented in Spectrum 6.2 in November 2022.
Previous mid-year interpolation is still available in code by using argument
create_fp(..., projection_period = "midyear"). The default argument
projection_period = NULL will choose the projection period based on the
Spectrum inputs version number. For versions 6.19 and earlier,
projection_period = "midyear". For versions 6.2 and later, projection_period = "calendar".
Normalise asfd and netmigagedist to sum to exactly 1.0 before distributing
to age groups.
Disaggregate under-5 net migration proportional to paediatric survival probabilities in the base year.
Net-migrations added at end of projection step, consistent with WPP 2022. No longer (1) adjust net migration for half-period survival, nor (2) adjust net migration to be half in current age group and half in next age group.
End-year ART input interpolation adjusted to reflect calendar year projection period, such that ART input aligns to end of projection year. Note: since net migration is applied after ART initiations, the number on ART will be scaled by the net migration proportions. Therefore the model output for number on ART will not exactly match the ART inputs.
In likelihood calculations for household survey proportion ever tested, interopolate
model outputs to mid-year. This affects ll_evertest(). Likelihood for annual
HIV tests and diagnoses (ll_prgdat()) is unaffected.
eppmod = "directinfections_hts" option (v1.5.0) to avoid referencing
uninitialised memory.Updates for 2022 UNAIDS estimates:
Extend output arrays, plots, and tables to output 2021 results.
Extend theta0 initial values for additional year random-walk parameters.
Intercalate new infections in each HIV time step. Equal number of infections are added in each time step. This is controlled by the option eppmod = "directinfections_hts".
Option eppmod = "directinfections_hts" is specified as the default option in
prepare_inputs(pjnz). Therefore, new package version will not reproduce simulations
from a previous model fit. Simulating from a previous fit requires manually specifying
fp$eppmod <- "directinfections" after calling prepare_inputs().
extract_pjnz() reads whether custom population adjustment was used in the Spectrum file from the tag "<RegionalAdjustPopCBState MV>". This is used to set popadjust = TRUE automatically if a custom population was used in Spectrum.
add_ss_indices(), add argument type.convert(..., as.is = TRUE)
to suppress R 4.0 warning.double incrate_g[NG]; in EPP_DIRECTINCID incidence
option. This did not affect any results because this option is not used in Shiny90 application;
new infections by sex and age group were directly specified (option EPP_DIRECTINFECTIONS).create_hts_param() will throw and error if an unexpected parameter
vector length is identified.Updates for 2021 UNAIDS estimates:
Extended arrays to output 2020 results.
Changed prior for baseline testing rate to normal(log(0.005), sd = 0.25). (likelihood.R#138)
Revise HIV retesting rate to increase log-linearly betwee 2005 and 2010. The implication of this is to reduce the retesting rate during the 2000s compared to the 2010s, which reduces implausibly high predictions for number of HIV tests conducted during the 2000s in the abscence of HTS programme data (Giguere et al 2020).
NEWS.md file to track changes to the package.