Modelling children in care demand and cost
Refer to the template config file at data/demo/config.edn. The config file specifies file inputs, output parameters and projection parameters.
There are eight standard input files, which are created by generating candidates and rejection sampling. Rejection sampling generates a variety of input period candidates together with their rejection sampling probabilities (rejection sampling is used to ensure that overall period durations are as historically observed).
The process is as follows:
- Execute
clj -m cic.cic run-generate-candidates-workflow
with a config to generate a period candidates file. - Execute
clj -m cic.cic run-rejection-sampling
with a config to generate model input files from the candidates file. - Execute
clj -m cic.cic run-cic-workflow
with a config to generate the projection. Depending on the config either or both of the episodes or summary csvs will be generated.
Alternatively:
- Execute
clj -m cic.main -main
optionally passing--config FILE
to use custom config.
- Total time in care - the total length of time in care
- Period - the length of contiguous time in care
- Phase - the length of time in a placement type
- Episode - the length of time spent in the same placement/placement provider/legal status/CIN
- CIN - acronym for 'child in need', however corresponds to category of need, being the main reason a child is being provided with care services.
The episodes output contains a provenance field which contains a single character representing the source of the period.
- H: Historic - historic joiners who leave before the projection start date
- O: Open - historic joiners who have not left by the projection start date
- P: Projected closed - Open periods the model closes
- C: Candidate - potential simulated joiners whose placements and age of entry is known but who do not have a start date
- S: Simulated - joiners the model creates by assigning a start date
The provenance is the same for all episodes in a period.