PR for issue 177 (simulation covariate prevalence)

[ob325]

Adds optional minCellCount parameter to createCohortMethodDataSimulationProfile. Default value is 10, with the idea it's better to protect the user from accidental disclosure of PHI. Adds non-exported function .truncateSimulationProfile that accepts a cohortDataSimulationProfile object and returns a modified copy having small cell counts reset to zero.

Example use:

x <- list()
> x$metaData <- list()
> x$metaData$populationSize <- 135000
> x$covariatePrevalence <- data.frame(covariateId = 1:1000, prevalence = runif(n = 1000, min = 0, max = .1))
> class(x) <- "CohortDataSimulationProfile"
> lapply(x, head)
$metaData
$metaData$populationSize
[1] 135000


$covariatePrevalence
  covariateId   prevalence
1           1 0.0005396071
2           2 0.0383829203
3           3 0.0393560038
4           4 0.0885898456
5           5 0.0140272565
6           6 0.0441554085

> 
> x0 <- .truncateSimulationProfile(x, 0)
Warning message:
In .truncateSimulationProfile(x, 0) :
  No truncation was done on low-prevalence covariates. Object may include low cell counts that enable identification of persons.
> identical(x, x0)
[1] TRUE
> x1 <- .truncateSimulationProfile(x, 100)
Before truncating simulation profile, lowest non-zero covariate prevalence is 0.00007826 (11 / 135000)
After truncating simulation profile, lowest non-zero covariate prevalence is 0.00077438 (105 / 135000)
> lapply(x1, head)
$metaData
$metaData$populationSize
[1] 135000


$covariatePrevalence
  covariateId prevalence
1           1 0.00000000
2           2 0.03838292
3           3 0.03935600
4           4 0.08858985
5           5 0.01402726
6           6 0.04415541

> 

[msuchard]

@ob325 -- default truncation across the OHDSI tool-stack appears to be 5 and not 10.

[ob325]

@msuchard just made new commit with default = 5