Error in creating a function to perform ttests on multiple continuous variables - r

Question

So I'm trying to create a function that will take in a string of continuous variables, a categorical variable and a dataframe and output a table that includes, for each continuous variable: mean group1, mean group2, teststat, confidence interval, p-value.
What is currently here gives me the error: Error in model.frame.default(formula = var ~ class, data = data) : variable lengths differ (found for 'class')
I would love any feedback on how to fix this error and make this function do what I like. I want to make this function way more substantial and flexible, but I can't even get the basic version (handling multiple variables) to work.
THANKS!
#Continuous must be an object of the form:
#vars<-c("cont1", "cont2", "cont3", etc)
#CREATE DATA
cat1<-sample(c(1,2), 100, replace=T)
cont1<-rnorm(100, 25, 8)
cont2<-rnorm(100, 0, 1)
cont3<-rnorm(100, 6, 14.23)
cont4<-rnorm(100, 25, 8)*runif(5, 0.1, 1)
one<-data.frame(cat1, cont1, cont2, cont3, cont4)
#FUNCTION
two.group.comp<-function(continvars,class,data){
attach(data)
descriptives<-function(var){
test<-t.test(var~class, data)
means<-data.frame(test[5])
mean1<-means[1,1]
mean2<-means[2,1]
teststatbig<-data.frame(test[1])
teststat<-teststatbig[1,1]
conf<-data.frame(test[4])
lconf<-conf[1,1]
uconf<-conf[2,1]
pvalues<-data.frame(test[3])
pvalue<-pvalues[1,1]
variablename<-deparse(substitute(var))
entry<-data.frame(variablename,mean1,mean2,lconf,uconf,teststat,pvalue)
}
var<-data.frame(continvars)
table<<-sapply(var,descriptives)
detach(data)
}
#VARIABLES
continvars<-c("cont1", "cont2", "cont3")
#CALL TO FUNCTION
two.group.comp(continvars=continvars, class=cat1, data=one)

Answer 1

Does this do what you want?
two.group.comp <- function(continvars,class,data){
get.stats <- function(x,cat){
f <- unique(cat)
x1 <- x[cat==f[1]]
x2 <- x[cat==f[2]]
tt <- t.test(x1,x2)
smry <- c(tt$estimate,tt$statistic,p=tt$p.value)
names(smry) <- c("mean.1","mean.2","t","p")
return(smry)
}
result <- do.call(rbind,lapply(data[,continvars],get.stats,cat=class))
return(result)
}
# create sample dataset
set.seed(1)
cat1 <-sample(c(1,2), 100, replace=T)
cont1<-rnorm(100, 25, 8)
cont2<-rnorm(100, 0, 1)
cont3<-rnorm(100, 6, 14.23)
cont4<-rnorm(100, 25, 8)*runif(5, 0.1, 1)
one <-data.frame(cat1, cont1, cont2, cont3, cont4)
continvars<-c("cont1", "cont2", "cont3")
# call the function...
two.group.comp(continvars,cat1,one)
# mean.1 mean.2 t p
# cont1 24.4223859 25.33275704 -0.6024497 0.54827955
# cont2 0.0330148 0.01168979 0.1013519 0.91947827
# cont3 10.5784201 4.00651493 2.4183031 0.01747468
Working from the inside out:
get.stats(...) takes a single column of data, splits it into x1 and x2 according to cat, runs the t-test, and returns the summary statistics as a named vector.
lapply(...) passes the continvars columns of data to get.stats(...) one at a time.
do.call(rbind,...) binds together the set of vectors returned from lapply(...), row-wise, to generate the final result table.
This will work also if you pass column numbers instead of column names.
A piece of advice: the way you have it set up, you pass the column names of the continuous variables, but you pass the grouping factor as a vector. It would be cleaner if you pass the column name of the grouping factor.