UACR/TI 90days - continuous N=1 - error

[romainkp]

The code below leads to this error:

Error in runglmMSM(OData, wts_data, all_dummies, Ynode, verbose) :
trying to get slot "model" from an object (class "try-error") that is not an S4 object

Note that the function g.Nstatic10001 below implies that the monitoring intervention is 11111 when number.0.between.1s <- 0

# ESTIMATE SURVIVAL FOR ONE TREATMENT REGIMEN WITH STATIC (1,0,0,0,1,0,...) INTERVENTIONS ON MONITORING REGIME
# get the likelihood for the following static g^* N(t): 1,0,0,0,1,0,0,0,1,0
g.Nstatic10001 <- function(){
  function(OdataDT, gn.N = "g0.N", SHIFTED.OUTCOME = "outcome.tplus1", ID = "ID", MONITOR = "N", t = "t", ...){
    ID.expression <- as.name(ID)
    Odata_sel <- OdataDT[, c(ID, MONITOR, t, SHIFTED.OUTCOME, gn.N), with = FALSE]
    number.0.between.1s <- 0 ## every k weeks means k-1 0's between 1's and this shoudl be set to k-1
    N.star.t <- as.integer((Odata_sel[[t]]%%(number.0.between.1s+1))%in%number.0.between.1s) 
    Odata_sel[, g.N := as.numeric(as.integer(Odata_sel[[MONITOR]]) == N.star.t)] 
    return(Odata_sel[["g.N"]])
  }
}
# Define N(t) rule followers under static N.g.star: (1,0,1,0) and a column to the observed data.table in OData:
g.star.N.follow <- g.Nstatic10001()
OData$dat.sVar[, N.star.stat10001 := g.star.N.follow(OData$dat.sVar, gn.N = "g0.N", SHIFTED.OUTCOME = "outcome.tplus1", ID = "StudyID", MONITOR = "N", t = "intnum")]

#1. Obtain weighted data sets by rule:
wts.St.d7 <- getIPWeights(OData, gstar_TRT = "new.d7", gstar_MONITOR = "N.star.stat10001")
wts.St.d7.5 <- getIPWeights(OData, gstar_TRT = "new.d7.5", gstar_MONITOR = "N.star.stat10001")
wts.St.d8 <- getIPWeights(OData, gstar_TRT = "new.d8", gstar_MONITOR = "N.star.stat10001")
wts.St.d8.5 <- getIPWeights(OData, gstar_TRT = "new.d8.5", gstar_MONITOR = "N.star.stat10001")
wts.all <- list(d7 = wts.St.d7, d7.5 = wts.St.d7.5, d8 = wts.St.d8, d8.5 = wts.St.d8.5)
print(object.size(wts.all), units = "MB")
wts.all <- rbindlist(wts.all)
wts.all <- wts.all[!is.na(cumm.IPAW) & !is.na(outcome.tplus1) & (cumm.IPAW > 0), ]
print(object.size(wts.all), units = "MB")

data.table::setthreads(20) # will help load data into h2o
# MSM for hazard with regular weights:
t.breaks.byquarter <- c(1:9)-1 ## need to change or will get error due to bins with no event when computing IC - need to add as bug on github
MSM.IPAW <- survMSM(OData, wts_data = wts.all, t_breaks = t.breaks.byquarter, use_weights = TRUE, est_name = "IPAW", getSEs = getSEs)
# MSM for hazard with truncated weights:
MSM.trunc <- survMSM(OData, wts_data = wts.all, t_breaks = t.breaks.byquarter, use_weights = TRUE, trunc_weights = 20, est_name = "IPAWtrunc", getSEs = getSEs)
# crude MSM for hazard without any weights:
MSM.crude <- survMSM(OData, wts_data = wts.all, t_breaks = t.breaks.byquarter, use_weights = FALSE, est_name = "crude", getSEs = getSEs)
# save(list = c("MSM.IPAW", "MSM.trunc", "MSM.crude"), file = "./MSMs.Rdata")

[osofr]

Have you tried non-h20 glm? It appears that the glm call is resulting in the error, which is what being captured by "try-error". Most likely explanation is due to 0 observations for some bins.

[romainkp]

You are correct. Glm works and shows that there are no events under any of
the rules and that is because we filled in the last N before an even with
0 artificially. I will instead fill it with 1.

We might want to add a check that the outcome in the final regression is
not constant at any given value and stop the program elegantly in such
cases. Not a priority

NOTICE TO RECIPIENT: If you are not the intended recipient of this
e-mail, you are prohibited from sharing, copying, or otherwise using or
disclosing its contents. If you have received this e-mail in error,
please notify the sender immediately by reply e-mail and permanently
delete this e-mail and any attachments without reading, forwarding or
saving them. Thank you.

From: Oleg Sofrygin [email protected]
To: osofr/stremr [email protected]
Cc: Romain S Neugebauer/CA/KAIPERM@KAIPERM, Author
[email protected]
Date: 06/24/2016 10:59 AM
Subject: Re: [osofr/stremr] UACR/TI 90days - continuous N=1 - error
(#12)

Caution: This email came from outside Kaiser Permanente. Do not open
attachments or click on links if you do not recognize the sender.

Have you tried non-h20 glm? It appears that the glm call is resulting in
the error, which what being captured. Most likely explanation is due to 0
observations for some bins.
—
You are receiving this because you authored the thread.
Reply to this email directly, view it on GitHub, or mute the thread.