OHDSI Home | Forums | Wiki | Github

Error in running estimation R code

(Gyeol Song) #1


I’m using ATLAS v2.5.0 and repeatedly getting this error message - Error in open.ff(x) : unable to open - when running a specific estimation R code locally.

traceback() returns the following log:

12: open.ff(x)
11: open(x)
10: `filename<-.ff`(`*tmp*`, value = ffcolname)
9: `filename<-`(`*tmp*`, value = ffcolname)
8: move.ffdf(x, dir = ".", name = n, relativepath = relativepath)
7: assign(n, move.ffdf(x, dir = ".", name = n, relativepath = relativepath))
6: ffbase::save.ffdf(covariates, covariateRef, analysisRef, dir = file, 
       clone = TRUE)
5: saveCohortMethodData(cohortMethodData, params$cohortMethodDataFolder)
4: FUN(X[[i]], ...)
3: lapply(x, fun, ...)
2: OhdsiRTools::clusterApply(cluster, objectsToCreate, createCmDataObject)
1: CohortMethod::runCmAnalyses(connectionDetails = connectionDetails, 
       cdmDatabaseSchema = cdmDatabaseSchema, exposureDatabaseSchema = exposureDatabaseSchema, 
       exposureTable = exposureTable, outcomeDatabaseSchema = outcomeDatabaseSchema, 
       outcomeTable = outcomeTable, cdmVersion = cdmVersion, outputFolder = outputFolder, 
       cmAnalysisList = cmAnalysisList, drugComparatorOutcomesList = drugComparatorOutcomesList, 
       getDbCohortMethodDataThreads = 1, createPsThreads = 1, psCvThreads = min(16, 
           maxCores), computeCovarBalThreads = min(3, maxCores), 
       createStudyPopThreads = min(3, maxCores), trimMatchStratifyThreads = min(10, 
           maxCores), fitOutcomeModelThreads = max(1, round(maxCores/4)), 
       outcomeCvThreads = min(4, maxCores), refitPsForEveryOutcome = FALSE)

Could anyone advise how I can get solve this error?

Thanks in advance.


(Gyeol Song) #2

Found the solution: don’t use Korean in the outputFolder directory!

(Kristin Kostka, MPH) #3

@gyeol99 I admire your tenacity. :slight_smile: I, too, reply to myself on the Forum when I find a solution.

I had an inkling about this being related to how it was reading in your outputFolder but I hadn’t had a cycle to think it through and reply. Cool! Well… good to know that R is inherently biased to Latin alphabets. :laughing:

Thanks for keeping at it!! If you encounter more issues, please keep asking questions. We are listening. @SCYou and I do read these and have lessons we can share.

(Gyeol Song) #4

I too replied myself hoping this thread may help other researchers in Korea facing the same problem.

Thanks @krfeeney for your feedback and encouragement.