issue with survfit()

[ohdsi]

issue with survfit()

[msuchard]

This is almost surely an int64 issue. I'll look into it. If you could provide a small code snippet that causes the error, that would be quite helpful.

[msuchard]

@jreps -- this following does cause an error (both in Cyclops with the first covariateId that I need to fix and in bit64 in representing the 2nd #). But, I think you are experiencing something different. Can you add to this code snippet?

test_that("Check int64 reductions", {
    set.seed(123)

    data <- simulateCyclopsData(nstrata = 1,
                                nrows = 1000,
                                ncovars = 5,
                                model="logistic")

    # using int64
    data$outcomes$rowId <- bit64::as.integer64(data$outcomes$rowId)
    data$covariates$rowId <- bit64::as.integer64(data$covariates$rowId)
    data$covariates$covariateId <- bit64::as.integer64(data$covariates$covariateId)
    data$covariates$covariateId[1] <- bit64::as.integer64( "9223372036854775807")
    data$covariates$covariateId[2] <- bit64::as.integer64("18446744073709551615")

    cyclopsData64 <- convertToCyclopsData(data$outcomes,data$covariates, modelType = "lr", addIntercept = TRUE)
    data$covariates$covariateId[1:3]
    summary(cyclopsData64)
})

[jreps]

I think the summary() issue with your data example is the same as in the summary() issue with the fitted model I saw yesterday:

set.seed(123)
  
  data <- simulateCyclopsData(nstrata = 1,
                              nrows = 1000,
                              ncovars = 5,
                              model="logistic")
  
  # using int64
  data$outcomes$rowId <- bit64::as.integer64(data$outcomes$rowId)
  data$covariates$rowId <- bit64::as.integer64(data$covariates$rowId)
  data$covariates$covariateId <- bit64::as.integer64(data$covariates$covariateId)
  data$covariates$covariateId[1] <- bit64::as.integer64( "9223372036854775807")
  data$covariates$covariateId[2] <- bit64::as.integer64("18446744073709551615")
  
  cyclopsData64 <- convertToCyclopsData(data$outcomes,data$covariates, modelType = "lr", addIntercept = TRUE)
  data$covariates$covariateId[1:3]
  
  fit <- Cyclops::fitCyclopsModel(cyclopsData64, prior = createPrior("laplace", useCrossValidation = TRUE)) 
  
  meanLinearPredictor(fit) # seems to give the same error 

[msuchard]

This commit (develop branch) should fix the issue (and probably introduce many new ones):

cdf3aa5