StepReg with missing imputation

I'm trying to do a stepwise Cox regression imputing missing values.

As a first step I tried a simple stepwise procedure using the StepReg package but I ran into the following issue:

set.seed(123) library(tidyverse) lung_na <- survival::lung %>% mutate(sex = factor(sex, levels = c(1, 2))) %>% # make sex as factor na.omit() res6 <- StepReg::stepwise( formula = survival::Surv(time, status) ~ age + sex + ph.ecog + ph.karno + pat.karno + meal.cal + wt.loss, data = lung_na, type = "cox", strategy = "forward", metric = "AICc" ) #> Error in data.frame(Parameter = c("initial formula", "regression type", : arguments imply differing number of rows: 11, 12 res6 #> Error: object 'res6' not found 

^{Created on 2025-01-03 with reprex v2.1.1}

I can't seem to understand why an error occurs using a complete formula and not using survival::Surv(time, status) ~ . as the formula inside the example.

I'm using the complete formula because the following step would be, using the mice package, this:

model_lung_imp <- with( data = lung_imp, StepReg::stepwise( formula = survival::Surv(time, status) ~ age + sex + ph.ecog + ph.karno + pat.karno + meal.cal + wt.loss, data = ???, type = "cox", strategy = "forward", metric = "AICc" ) ) 

which doesn't work with dot formulas. In this case it's also not particularly clear how to pass the data argument to the with() formula.

This is an example with a normal Cox regression:

set.seed(123) library(survival) library(mice) #> #> Attaching package: 'mice' #> The following object is masked from 'package:stats': #> #> filter #> The following objects are masked from 'package:base': #> #> cbind, rbind lung_imp <- survival::lung |> mice::mice(method = "cart", print = F) model_lung_imp <- with( data = lung_imp, survival::coxph(survival::Surv(time, status) ~ age + sex + ph.ecog + ph.karno + pat.karno + meal.cal + wt.loss) ) summary(mice::pool(model_lung_imp)) #> term estimate std.error statistic df p.value #> 1 age 1.166524e-02 0.0095162812 1.22581919 154.84482 0.2221277796 #> 2 sex -5.918487e-01 0.1717995858 -3.44499507 155.11338 0.0007347178 #> 3 ph.ecog 5.728645e-01 0.1870627320 3.06241937 145.77603 0.0026153061 #> 4 ph.karno 1.341023e-02 0.0094293986 1.42217219 154.85166 0.1569874005 #> 5 pat.karno -1.269711e-02 0.0070854108 -1.79200694 137.03091 0.0753384836 #> 6 meal.cal -1.991951e-05 0.0002478896 -0.08035637 66.20331 0.9361962526 #> 7 wt.loss -1.003045e-02 0.0066743084 -1.50284529 127.96469 0.1353428021 

^{Created on 2025-01-03 with reprex v2.1.1}

Any tips on how to accomplish this?

Thank you.