Publish (version 2023.01.17)

publish.MIresult: Present logistic regression and Cox regression obtained with mitools::MIcombine based on smcfcs::smcfcs multiple imputation analysis

Description

Regression tables after multiple imputations

Usage

# S3 method for MIresult
publish(
  object,
  confint.method,
  pvalue.method,
  digits = c(2, 4),
  print = TRUE,
  factor.reference = "extraline",
  intercept,
  units = NULL,
  fit,
  data,
  ...
)

Arguments

object

Object obtained with mitools::MIcombine based on smcfcs::smcfcs multiple imputation analysis

confint.method

No options here. Only Wald type confidence intervals.

pvalue.method

No options here. Only Wald type tests.

digits

Rounding digits for all numbers but the p-values.

print

If FALSE suppress printing of the results

factor.reference

Style for showing results for categorical. See regressionTable.

intercept

See regressionTable.

units

See regressionTable.

fit

One fitted model using the same formula as object. This can be the fit to the complete case data or the fit to one of the completed data. It is used to get xlevels, formula and terms. For usage see examples. is used to fit

data

Original data set which includes the missing values

...

passed to summary.regressionTable, labelUnits and publish.default.

Author

Thomas A. Gerds <[e-mail]>

Details

Show results of smcfcs based multiple imputations of missing covariates in publishable format

Examples

if (FALSE) {
if (requireNamespace("riskRegression",quietly=TRUE)
  & requireNamespace("mitools",quietly=TRUE)
  & requireNamespace("smcfcs",quietly=TRUE)){
library(riskRegression)
library(mitools)
library(smcfcs)
## continuous outcome: linear regression
# lava some data with missing values
set.seed(7)
d=sampleData(78)
## generate missing values
d[X1==1,X6:=NA] 
d[X2==1,X3:=NA]
d=d[,.(X8,X4,X3,X6,X7)]
sapply(d,function(x)sum(is.na(x)))

# multiple imputation (should set m to a large value)

set.seed(17)
f= smcfcs(d,smtype="lm",
           smformula=X8~X4+X3+X6+X7,
           method=c("","","logreg","norm",""),m=3)
ccfit=lm(X8~X4+X3+X6+X7,data=d)
mifit=MIcombine(with(imputationList(f$impDatasets),
                lm(X8~X4+X3+X6+X7)))
publish(mifit,fit=ccfit,data=d)
publish(ccfit)

## binary outcome
# lava some data with missing values
set.seed(7)
db=sampleData(78,outcome="binary")
## generate missing values
db[X1==1,X6:=NA] 
db[X2==1,X3:=NA]
db=db[,.(Y,X4,X3,X6,X7)]
sapply(db,function(x)sum(is.na(x)))

# multiple imputation (should set m to a large value)
set.seed(17)
fb= smcfcs(db,smtype="logistic",
           smformula=Y~X4+X3+X6+X7,
           method=c("","","logreg","norm",""),m=2)
ccfit=glm(Y~X4+X3+X6+X7,family="binomial",data=db)
mifit=MIcombine(with(imputationList(fb$impDatasets),
                glm(Y~X4+X3+X6+X7,family="binomial")))
publish(mifit,fit=ccfit)
publish(ccfit)

## survival: Cox regression
library(survival)
# lava some data with missing values
set.seed(7)
ds=sampleData(78,outcome="survival")
## generate missing values
ds[X5==1,X6:=NA] 
ds[X2==1,X3:=NA]
ds=ds[,.(time,event,X4,X3,X6,X7)]
sapply(ds,function(x)sum(is.na(x)))

set.seed(17)
fs= smcfcs(ds,smtype="coxph",
           smformula="Surv(time,event)~X4+X3+X6+X7",
           method=c("","","","logreg","norm",""),m=2)
ccfit=coxph(Surv(time,event)~X4+X3+X6+X7,data=ds)
mifit=MIcombine(with(imputationList(fs$impDatasets),
                coxph(Surv(time,event)~X4+X3+X6+X7)))
publish(mifit,fit=ccfit,data=ds)
publish(ccfit)

## competing risks: Cause-specific Cox regression 
library(survival)
# lava some data with missing values
set.seed(7)
dcr=sampleData(78,outcome="competing.risks")
## generate missing values
dcr[X5==1,X6:=NA] 
dcr[X2==1,X3:=NA]
dcr=dcr[,.(time,event,X4,X3,X6,X7)]
sapply(dcr,function(x)sum(is.na(x)))

set.seed(17)
fcr= smcfcs(dcr,smtype="compet",
           smformula=c("Surv(time,event==1)~X4+X3+X6+X7",
                       "Surv(time,event==2)~X4+X3+X6+X7"),
           method=c("","","","logreg","norm",""),m=2)
## cause 2 
ccfit2=coxph(Surv(time,event==2)~X4+X3+X6+X7,data=dcr)
mifit2=MIcombine(with(imputationList(fcr$impDatasets),
                coxph(Surv(time,event==2)~X4+X3+X6+X7)))
publish(mifit2,fit=ccfit2,data=dcr)
publish(ccfit2)
}
}