Recover exact expected outcomes from a simulated panel

Keeps count-level effects and log mean ratios distinct. This function deliberately does not treat a log mean as the expectation of log counts.

Usage

lamp_simulation_truth(
  panel,
  outcome = c("affected_crime", "crime_total"),
  scale = c("identity", "log_mean")
)

Arguments

panel

A panel returned by lamp_simulate().

outcome

"affected_crime" or "crime_total".

scale

"identity" for expected count differences or "log_mean" for log ratios of expected counts.

Value

A tibble of area-month counterfactual means and exact effects, including cohort, relative time and observed coverage status. For spillover designs the effect includes the simulated neighbour exposure.

See also

Other simulation: lamp_simulate()

Examples

sim <- lamp_simulate(seed = 1)
head(lamp_simulation_truth(sim))
#> # A tibble: 6 × 11
#>   area   month      cohort rel_time treated coverage_status mean_untreated
#>   <chr>  <date>      <int>    <int>   <dbl> <chr>                    <dbl>
#> 1 S00001 2018-01-01      9       -8       0 submitted                 6.32
#> 2 S00001 2018-02-01      9       -7       0 submitted                 6.63
#> 3 S00001 2018-03-01      9       -6       0 submitted                 6.77
#> 4 S00001 2018-04-01      9       -5       0 submitted                 6.69
#> 5 S00001 2018-05-01      9       -4       0 submitted                 6.42
#> 6 S00001 2018-06-01      9       -3       0 submitted                 6.07
#> # ℹ 4 more variables: mean_treated <dbl>, effect <dbl>, outcome <chr>,
#> #   scale <chr>