Skip to content

Latest commit

 

History

History
106 lines (86 loc) · 5.07 KB

NOTES.md

File metadata and controls

106 lines (86 loc) · 5.07 KB

DONE:

  1. EUROPE => EUROPE-NORTH AMERICA
  2. Rename "UNK" to "OTHER" in "Region For Migration.." and combine with "CARIBBEAN..." for migration modelling
  3. Add info that records with "UNK" are removed from the dataset processed for migration
  4. Reverse order of the X-axis
  5. Keep of order conditions in PrepareMigrantData
  6. Show count of cases with missing transmission category
  7. Add info about combining CARLAM and OTHER
  8. Add note that date of diagnosis is the date of diagnosis in the reporting country
  9. Check selection of attributes. Currently the focused option is selected.
  10. Migrant diagnosis - Show info that we show average and round to the nearest integer
  11. Incorporate HIVStatus into KnownPrePost
  12. Add tab in "Modelling" -> "Migrant connection" (before "Populations") and enable if migration module was run
  13. Make migration connection optional in the modelling module
  14. If migrant connection is selected then aggregated data sets override is deselected and inactive, except Dead.
  15. Check why there are missings in GroupedRegionOfOrigin.
  16. Add three new columns to the outputs of the model, rather than adjusting existing columns
  17. Create a new column that is a composite of stratification variables.

TODO:

  1. Derive the incidence curve as exp(average(log(BSpline)) over adjusted data set, get average diagnosis rate matrix and compute all HIV model outputs. Use that as the "benchmark" HIV model fit for plots, tables.
  2. Research comparison between main fit point estimates and the median of bootstrap
  3. If small data set issue a warning that there are too few cases to estimate confidence bounds. Show median as point estimate instead, no bounds.
  4. Filter results on length of betas - keep only those that have the length of betas consistent with the length of the levels of factor. Issue a warning when this happens (results are based on a lower number of imputations).
  5. Reformat data to have Imputation count times the sample size. Drop incomplete data (strat), (max 10%). Always print out the percent of sample used.
  6. Test splines for year of arrival

// 2022/05/09

  1. Migrant conf: check in CheckAggregated which strata has PresentRatio < 0.9 and remove them from the data. Compute GLM based on that filtered dataset. Base the probabilities for the removed categories on mean of ProbPre.
  2. HIV modelling combining: thetas can have different length if models got simplified. Allow passing spline as vector to cpp side, rather than having it computed their. Pass average spline, rather than average theta (since we can't have that).

// 2022/05/20

Output charts:

  1. Migrants who arrive infected in chart "HIV infections per year", title "New arrivals of infected migrants"
  2. Chart 2 - time between arrival and infection
  3. Chart 3 - show 2 curves and total, add count of undiagnosed and diagnosed in this chart just like in the Windows version.
  4. Chart 4 - Fix y-axis title ("Proportion")

Colors of curves:

  • green - output of the model
  • blue - migration
  • black - total

// 2022/06/09

  1. Chart A.

    N_Inf_M: New incident infections NewMigrantDiagnosesPerArrYear: New arrivals of infected migrants Total: N_Inf_M + NewMigrantDiagnosesPerArrYear

    Add mapping from model outputs to chart labels

  2. Chart B. No changes

  3. Chart C.

    Diagnosed from model: N_Alive_Diag_M Diagnosed migrants: CumDiagnosedCasesInclMigr Undiagnosed from model: N_Und Undiagnosed migrants: CumInfectionsInclMigr - CumDiagnosedCasesInclMigr Alive: sum of all the above

  4. Chart D.

    if migrant module: UndiagnosedFrac else N_Und_Alive_p

  5. Enable reporting after adjustments.

  6. Allow hiding confidence bounds in charts (establish count criteria for default setting).

  7. Add count of observations in chart tooltip (per point).

  8. Check how mortality is integrated with non-parametric main model fit.

// 2022/11/03

  • Allow unloading data (both case-based and aggregated)
  • Move migrant connection to "Main fit run"
  1. Prepare version 3.0.0 (DONE)
  2. Build R packages and move to ftp server (DONE)
  3. Build Windows deployment package with R 4.2.2 (DONE)
  4. Create a presentation with technical details of the tool (DONE)
  5. Record a video:
    • Describe technology used - R (server) and JavaScript (UI)
  • Describe core R packages (hivModelling, hivPlatform)
  • Describe ways of accessing the tool: ECDC server; run Windows deployment tool; install and run R package
  1. Check all steps of calculations, most notably reporting delays with trend (in progress)
  2. Implement unloading data
  3. Implement saving/loading state of the entire application
  4. Align styling of logs printed
  5. Check behaviour of the tool when run as installed package, not as loaded with pkgload. (DONE)
  6. Change deployment on ECDC server to use installed package, not code loaded with pkgload. (DONE)
  7. Create CLI example to showcase the tool when running from console.
  8. Fix github action in hivModelling repo. (DONE)
  9. Move package dependencies management from renv to pak (DONE)
  10. Number of mice iterations: 10 (DONE)
  11. Fix reporting with MI and RD adjustments (DONE)
  12. Rotate labels in MI diagnostic plots