Error after first PQL iteration (iteration0) for MixRFb

[pietjonker]

Dear Jiebiao,

I am receiving on of two errors for 10 of the 12 models I want to run using your function

Error in RET@prediction_weights(newdata = newdata, mincriterion = mincriterion, : cannot compute out-of-bag predictions for observation number 6 | Error in model@dpp(...). : missing values in response variable not allowed

when running the MixRFb function.

I have a binomial (0 or 1) outcome variable (let's call this y1) and four predictor variables. The weird thing is that it works on another similar outcome variable (let's call this y2). I do exactly the same, the data is the same, only the prevalence of 1's is different. y1 (which doesn't work) has low prevalence of 1's and y2 has okay prevalence of 1's. Could this be the problem?

I have a total of 12 y variables for which only 4 run the full iterations, and 8 run into one of these two errors after the first iterations0.

Could you help me in the right direction? Thanks for the package and the help!!

[randel]

Dear Piet,

For your second error, Error in model@dpp(...). : missing values in response variable not allowed, are there any missing values in your response variables?

The first error is related to the underlying cforest function. Not sure if the link below helps. https://stackoverflow.com/questions/49893104/issue-with-caseweights-and-oob-prediction-using-cforest-from-the-party-package

How low is the prevalence of 1's? It's hard to say without seeing a replicated data example. You may email me if you hesitate to share a subset of data publicly.

[pietjonker]

Dear Jiebiao,

Thanks a lot for your response! I haven't found the answer in the stackoverflow page unfortunately. Sometimes I get the error even when the prevalence is around 50%.

For the NA error, I have checked multiple times. My data doesn't contain any missing values. If it does, I get an error immediatly in the first iteration. Now, I get an error for a missing value after a couple of iterations.

Note sure what is happening! Could it be that the bootstrapped RF sample only contains 0 outcomes sometimes, then throwing the error?

Kind regards, Piet

---

Van: Jiebiao Wang @.> Verzonden: vrijdag 23 april 2021 00:03 Aan: randel/MixRF @.> CC: pietjonker @.>; Author @.> Onderwerp: Re: [randel/MixRF] Error after first PQL iteration (iteration0) for MixRFb (#7)

Dear Piet,

For your second error, Error in @.***(...). : missing values in response variable not allowed, are there any missing values in your response variables?

The first error is related to the underlying cforest function. Not sure if the link below helps. https://stackoverflow.com/questions/49893104/issue-with-caseweights-and-oob-prediction-using-cforest-from-the-party-package

How low is the prevalence of 1's? It's hard to say without seeing a replicated data example. You may email me if you hesitate to share a subset of data publicly.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHubhttps://github.com/randel/MixRF/issues/7#issuecomment-825213344, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AIJKBPG572T6V64TRS4V6JDTKCMMZANCNFSM43KK77TQ.