mcfDiff() Error: Error in if (max_res_ij > 0) { : missing value where TRUE/FALSE needed

[alexhillcoat]

Hello,

First of all, thank you very much for this package! It has been extremely helpful in conducting my thesis.

I am trying to conduct a statistical test of the difference between two groups' MCF, both of which are in one MCF object. I had no issues in creating or plotting the object itself and looking at the object, there are no missing values or extreme values. However, when I try to use the mcfDiff() function on the object, I get the following warnings before it aborts with the error in the title:

Warning: longer object length is not a multiple of shorter object length
Warning: longer object length is not a multiple of shorter object length
Warning: longer object length is not a multiple of shorter object length
Warning: longer object length is not a multiple of shorter object length
Warning: longer object length is not a multiple of shorter object length
Error in if (max_res_ij > 0) { : missing value where TRUE/FALSE needed

I cannot share my data, but below is my code. tstop and tstart are integers representing the stop and start times respectively per row, status is a 1/0 indicator of event, miss_sexmin is a 1/0 indicator of group, and dnweight is a column representing with the sampling weights.

mcf_sens <- mcf(Recur(time=tstop, id=xwaveid, event=status, origin=tstart, check = "soft") ~ miss_sexmin, data=sensdat, weights=as.numeric(dnweight))
plot(mcf_sens, conf.int=T, addOrigin=T)
mcfDiff(mcf_sens)

Thank you in advance.

[wenjie2wang]

Thanks for reporting this issue. I will investigate when I get a chance.

[wenjie2wang]

My best guess is that some xwaveid inside sensdat contains multiple censoring time points where status <= 0. Did you get any warning message from the function Recur()?

[alexhillcoat]

Hi Wenjie,

There are not multiple censoring times, but Recur() does warn me that some subjects have an event at censoring. That is simply a feature of my data, though - some participants reported a (non-terminal) event at their last follow-up time. Do I have to remove these events?

Thank you in advance!

[wenjie2wang]

I cannot reproduce the issue with the following example, where I set the censoring time to be the last event time if there are multiple events. The function Recur() should not throw out any error/warning messages because of that either. The message you got from Recur() may suggest other issues of the input data. There is perhaps not much I can do without a reproducible example.

library(data.table)
library(reda)

test_dat <- as.data.table(reda::simuDat)
## set censoring at the last event time
test_dat[, right := if (.N > 2) {
                        c(head(time, -2), rep(time[.N], 2))
                    } else { time }, keyby = ID]
test_dat[, left := c(0, head(right, -1)), keyby = ID]
test_dat[, .(ID, left, right, event)]
#>       ID left right event
#>   1:   1    0     1     1
#>   2:   1    1    22     1
#>   3:   1   22    23     1
#>   4:   1   23   112     1
#>   5:   1  112   112     0
#>  ---                     
#> 496: 100  119   123     1
#> 497: 100  123   124     1
#> 498: 100  124   134     1
#> 499: 100  134   140     1
#> 500: 100  140   140     0

mcfDiff(mcf(Recur(left %to% right, ID, event, check = "soft") ~ group,
            data = test_dat))
#> Call: 
#> mcfDiff(mcf1 = mcf(Recur(left %to% right, ID, event, check = "soft") ~ 
#>     group, data = test_dat))
#> 
#> Two-Sample Pseudo-Score Tests:
#>                 Statistic Variance    Chisq DF Pr(>Chisq)   
#> Constant Weight   52.3947 669.5665   4.1000  1   0.042884 * 
#> Linear Weight     36.8567 157.2041   8.6411  1   0.003287 **
#> ---
#> Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
#> 
#> Variance Estimator: robust

^{Created on 2024-02-20 with reprex v2.1.0}

[alexhillcoat]

The error was (as expected) 100% on my end - I did not realize that I needed to add a row for people who had an event at their last observed time to indicate that they were also censored at that time, which I see from your example is necessary.

mcfDiff() is now working as expected for me!

Thank you very much for your patience and time, apologies that I did not catch it myself.

[wenjie2wang]

Glad to hear that you figured out the issue.