Modifying a R for loop to work with three variables - r

Question

I've got a dataset that i need to remove outliers from. The data is from a repeated measures experiment with 105 subjects. The code i am using only removes outliers from one of the conditions in the study ("markLiturEndurt") but not the other two ("markStadsEndurt" and "litStaEndurt"). Each condition is binary and has been coded into "Skilyrdi"
The code I am using looks like this:
(Skilyrdi<-unique(gogn$markLiturEndurt))
(fjoldiRada<-length(gogn$subject))
(fjoldiSkil<-length(Skilyrdi))
gognHrein<-0
for (i in 1:length(Skilyrdi))
{
gognSkil<-subset(gogn,gogn$markLiturEndurt==Skilyrdi[i])
numerThatt<-unique(gognSkil$subject)
for (j in 1:length(numerThatt))
{
gognThatt<-subset(gognSkil,gognSkil$subject==numerThatt[j])
(medalST<-mean(gognThatt$areitiRT))
(sfST<-sd(gognThatt$areitiRT))
(nedriMork<-100)
(efriMork<-medalST+3*sfST)
gognThatt<-subset(gognThatt,gognThatt$areitiRT>nedriMork)
gognThatt<-subset(gognThatt,gognThatt$areitiRT<efriMork)
gognHrein<-rbind(gognHrein,gognThatt)
}
}
I've tried messing around with the code but my limited R programming skills got the better of me. I'm pretty sure it should be possible either to modify the my code so that it checks for outliers in all the conditions or that maybe i could create additional for loops for each condition and then bind them together.
subject umferd areitiACC areitiRT markLiturEndurt markStadsEndurt litStaEndurt
117 1 1 1202 0 0 0
117 2 1 924 0 1 0
117 4 1 1139 0 0 0
117 5 1 1211 0 0 0
117 6 1 998 1 1 0
117 7 1 778 0 1 0
"areitiRT" is reaction time the dependent variable. "umferð" is each round, "areitiACC" is correct or wrong answer. The other three variables are the independent variables.
What i would like to achieve:
The depentant variable of interest for me is areitiRT this is "reaction time". The indepentant variables I am working with are markLiturEndurt markStadsEndurt litStaEndurt. Each indepentant variable can take two values "0" (IV changed) and "1" (IV held constant).
I want to see how reaction time differs under each condition that the DV can take by doing an ANOVA. Before i can do the ANOVA i need to clean the data so it does not contain any extreme values. Reaction times areitiRT under each condition of the DV for each subject need to be check for every observation under that condition and compared to that subjects mean reaction time. Reaction times less then 100ms and greater then the "mean + 3 * Standard Deviation" need to be removed.

Answer 1

It's taken me a week to get back to you because I've been on holiday, so sorry about the delay.
I'm still not entirely sure what you need, so here are some options. If you clarify what you need by editing your question I can edit my answer appropriately.
Example dataset
Use the following dataset as gogn as an example:
gogn <- structure(list(subject = c(117L, 117L, 117L, 117L, 117L, 117L,
117L, 118L, 118L, 118L, 118L, 118L, 118L), umferd = c(1L, 2L,
4L, 5L, 6L, 7L, 7L, 1L, 2L, 4L, 5L, 6L, 7L), areitiACC = c(1L,
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L), areitiRT = c(1202L,
924L, 1139L, 1211L, 998L, 778L, 53L, 1202L, 924L, 1139L, 1211L,
1024L, 778L), markLiturEndurt = c(0L, 0L, 0L, 0L, 1L, 0L, 0L,
0L, 0L, 0L, 0L, 1L, 0L), markStadsEndurt = c(0L, 1L, 0L, 0L,
1L, 1L, 1L, 0L, 1L, 0L, 0L, 1L, 1L), litStaEndurt = c(0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L)), .Names = c("subject",
"umferd", "areitiACC", "areitiRT", "markLiturEndurt", "markStadsEndurt",
"litStaEndurt"), class = "data.frame", row.names = c(NA, -13L
))
Remove reaction times <100ms
gogn <- gogn[gogn$areitiRT > 99, ]
Or, if you want to use dplyr:
require("dplyr")
gogn <- subset(gogn, areitiRT > 99)
In both cases this will remove reaction times less than 100ms (i.e. 100ms or greater).
Reaction times 3 * sd + mean
This is a bit more involved, but not too difficult. First, we need a list of all unique subjects:
uSubject <- as.list(unique(gogn$subject))
Then we create a list which contains the value of 3 * sd + mean for each subject:
outliers <- lapply(uSubject, function(x){
3 * sd(gogn$areitiRT[gogn$subject == x]) +
mean(gogn$areitiRT[gogn$subject == x])
})
Then I prefer to input this back in to the data frame (gogn) as a new column for easy use:
for(i in 1:length(outliers)){
gogn$outlier[gogn$subject == uSubject[i]] <- outliers[i]
}
Then we can compare each individual reaction time with the 'outlier value' we just calculated and, if it's greater, ditch it:
gogn <- gogn[which(gogn$areitiRT < gogn$outlier), ]
# this removes any cases where reaction time is NOT less than the limit
Is this answer along the lines of what you need to do?