andreas/daDoctoR
1
0
Fork 0
mirror of https://github.com/agdamsbo/daDoctoR.git synced 2024-11-21 19:30:22 +01:00
Code Issues Projects Releases Wiki Activity

updated strobe_pred

Browse Source 
still needs flag to include p-values
This commit is contained in:
agdamsbo 2019-11-11 17:16:25 +01:00
parent 8326ed2cc0
commit cc24e7d209
2 changed files with 162 additions and 8 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

166
R/strobe_pred.R
View File

@ -1,6 +1,6 @@
#' Logistic regression of predictors according to STROBE
#' Regression model of predictors according to STROBE, bi- and multivariate.
#'
#' Printable table of logistic regression analysis according to STROBE.
#' Printable table of regression model according to STROBE. Includes borth bivariate and multivariate in the same table. Output is a list, with the first item being the main "output" as a dataframe. Automatically uses logistic regression model for dichotomous outcome variable and linear regression model for continous outcome variable.
#' @param meas binary outcome meassure variable, column name in data.frame as a string. Can be numeric or factor. Result is calculated accordingly.
#' @param adj variables to adjust for, as string.
#' @param data dataframe of data.
@ -21,6 +21,8 @@ strobe_pred<-function(meas,adj,data,dec=2,n.by.adj=FALSE){
ads<-d[,c(adj)]
if(is.factor(m)){
## Crude ORs
dfcr<-data.frame(matrix(NA,ncol = 3))
@ -36,11 +38,11 @@ strobe_pred<-function(meas,adj,data,dec=2,n.by.adj=FALSE){
n.mn<-c(n.mn,nrow(mn$model))
suppressMessages(ci<-exp(confint(mn)))
l<-round(ci[-1,1],2)
u<-round(ci[-1,2],2)
or<-round(exp(coef(mn))[-1],2)
l<-round(ci[-1,1],dec)
u<-round(ci[-1,2],dec)
or<-round(exp(coef(mn))[-1],dec)
or_ci<-paste0(or," (",l," to ",u,")")
pv<-round(tidy(mn)$p.value[-1],3)
pv<-round(tidy(mn)$p.value[-1],dec+1)
x1<-ads[,i]
if (is.factor(x1)){
@ -166,6 +168,158 @@ strobe_pred<-function(meas,adj,data,dec=2,n.by.adj=FALSE){
ls<-list(tbl=ref,miss,n.meas,nrow(d))
names(ls)<-c("Printable table","Deleted due to missingness in adjusted analysis","Number of outcome observations","Length of dataframe")
}
if(!is.factor(m)){
d<-dta
m<-d[,c(meas)]
ads<-d[,c(adj)]
dfcr<-data.frame(matrix(NA,ncol = 3))
names(dfcr)<-c("pred","mean_ci","pv")
n.mn<-c()
nref<-c()
for(i in 1:ncol(ads)){
dat<-data.frame(m=m,ads[,i])
names(dat)<-c("m",names(ads)[i])
mn<-lm(m~.,data=dat)
n.mn<-c(n.mn,nrow(mn$model))
suppressMessages(ci<-confint(mn))
l<-round(ci[-1,1],dec)
u<-round(ci[-1,2],dec)
mean<-round(coef(mn)[-1],dec)
mean_ci<-paste0(mean," (",l," to ",u,")")
pv<-round(tidy(mn)$p.value[-1],dec+1)
x1<-ads[,i]
if (is.factor(x1)){
pred<-paste0(names(ads)[i],levels(x1)[-1])
}
else {
pred<-names(ads)[i]
}
dfcr<-rbind(dfcr,cbind(pred,mean_ci,pv))
}
## Mutually adjusted ORs
dat<-data.frame(m=m,ads)
ma <- lm(m ~ ., data = dat)
miss<-length(ma$na.action)
actable <- coef(summary(ma))
pa <- actable[,4]
pa<-ifelse(pa<0.001,"<0.001",round(pa,3))
pa <- ifelse(pa<=0.05|pa=="<0.001",paste0("*",pa),
ifelse(pa>0.05&pa<=0.1,paste0(".",pa),pa))
apv<-pa[1:length(coef(ma))]
aco<-round(coef(ma),dec)
suppressMessages(aci<-round(confint(ma),dec))
alo<-aci[,1]
aup<-aci[,2]
amean_ci<-paste0(aco," (",alo," to ",aup,")")
nq<-c()
if (n.by.adj==TRUE){
dat2<-ma$model[,-1]
for (i in 1:ncol(dat2)){
if (is.factor(dat2[,i])){
vec<-dat2[,i]
ns<-names(dat2)[i]
for (r in 1:length(levels(vec))){
vr<-levels(vec)[r]
n<-as.numeric(length(vec[vec==vr&!is.na(vec)]))
nall<-as.numeric(length(dat2[,c(ns)]))
n.meas<-nall
nl<-paste0(ns,levels(vec)[r])
pro<-round(n/nall*100,0)
rt<-paste0(n," (",pro,"%)")
nq<-rbind(nq,cbind(nl,rt))
}}
if (!is.factor(dat2[,i])){
num<-dat2[,i]
nl<-names(dat2)[i]
n<-as.numeric(length(num[!is.na(num)]))
nall<-as.numeric(nrow(dat2))
n.meas<-nall
pro<-round(n/nall*100,0)
rt<-paste0(n," (",pro,"%)")
nq<-rbind(nq,cbind(nl,rt))
}}}
else {
dat2<-dat[!is.na(dat[,1]),][,-1]
n.meas<-nrow(dat2)
for (i in 1:ncol(dat2)){
if (is.factor(dat2[,i])){
vec<-dat2[,i]
ns<-names(dat2)[i]
for (r in 1:length(levels(vec))){
vr<-levels(vec)[r]
n<-as.numeric(length(vec[vec==vr&!is.na(vec)]))
nall<-as.numeric(n.mn[i])
nl<-paste0(ns,levels(vec)[r])
pro<-round(n/nall*100,0)
rt<-paste0(n," (",pro,"%)")
nq<-rbind(nq,cbind(nl,rt))
}}
if (!is.factor(dat2[,i])){
num<-dat2[,i]
nl<-names(dat2)[i]
n<-as.numeric(length(num[!is.na(num)]))
nall<-as.numeric(n.meas)
pro<-round(n/nall*100,0)
rt<-paste0(n," (",pro,"%)")
nq<-rbind(nq,cbind(nl,rt))
}}}
rnames<-c()
for (i in 1:ncol(dat2)){
if (is.factor(dat2[,i])){
rnames<-c(rnames,names(dat2)[i],paste0(names(dat2)[i],levels(dat2[,i])))
}
if (!is.factor(dat2[,i])){
rnames<-c(rnames,paste0(names(dat2)[i],".all"),names(dat2)[i])
}
}
res<-cbind(amean_ci,apv)
rest<-data.frame(names=row.names(res),res,stringsAsFactors = F)
numb<-data.frame(names=nq[,c("nl")],N=nq[,c("rt")],stringsAsFactors = F)
namt<-data.frame(names=rnames,stringsAsFactors = F)
coll<-left_join(left_join(namt,numb,by="names"),rest,by="names")
header<-data.frame(matrix("Adjusted",ncol = ncol(coll)),stringsAsFactors = F)
names(header)<-names(coll)
df<-data.frame(rbind(header,coll),stringsAsFactors = F)
names(dfcr)[1]<-c("names")
suppressWarnings(re<-left_join(df,dfcr,by="names"))
ref<-data.frame(re[,1],re[,2],re[,5],re[,3])
names(ref)<-c("Variable",paste0("N=",n.meas),"Crude OR (95 % CI)","Mutually adjusted OR (95 % CI)")
ls<-list(tbl=ref,miss,n.meas,nrow(d))
names(ls)<-c("Printable table","Deleted due to missingness in adjusted analysis","Number of outcome observations","Length of dataframe")
}
return(ls)
}

4
man/strobe_pred.Rd
View File

@ -2,7 +2,7 @@
% Please edit documentation in R/strobe_pred.R
\name{strobe_pred}
\alias{strobe_pred}
\title{Logistic regression of predictors according to STROBE}
\title{Regression model of predictors according to STROBE, bi- and multivariate.}
\usage{
strobe_pred(meas, adj, data, dec = 2, n.by.adj = FALSE)
}
@ -18,6 +18,6 @@ strobe_pred(meas, adj, data, dec = 2, n.by.adj = FALSE)
\item{n.by.adj}{flag to indicate wether to count number of patients in adjusted model or overall for outcome meassure not NA.}
}
\description{
Printable table of logistic regression analysis according to STROBE.
Printable table of regression model according to STROBE. Includes borth bivariate and multivariate in the same table. Output is a list, with the first item being the main "output" as a dataframe. Automatically uses logistic regression model for dichotomous outcome variable and linear regression model for continous outcome variable.
}
\keyword{logistic}