Subset function does not evaluate arguments - r

Question

Using the below data frame called 'data', I'm able to directly assign values to two variables, 'state' and 'measure', and identify the school with the lowest score within the subset:
Create dataframe 'data':
school<-c("NYU", "BYU", "USC", "FIT", "Oswego","UCLA","USF","Columbia")
state<-c("NY","UT","CA","NY","NY","CA", "CA","NY")
measure<-c("MSAT","MSAT","GPA","MSAT","MSAT","GPA","GPA","GPA")
score<-c(590, 490, 2.9, 759, 550, 1.2, 3.1, 3.2)
data<-data.frame(school,state, measure,score)
Subset on a 'state' and 'measure':
answer<-subset(data,subset=(state=="NY" & measure=="MSAT"))
order.answer<-order(answer$score,answer$school) #answer$school is tie-breaker
answer1<-as.matrix(answer[order.answer,])
answer1[1,1]
This is the correct answer:
[1] "Oswego"
My problem is that when I create a function to accomplish the same thing, I get an incorrect result:
lowest <- function(state, measure){
answer<-subset(data,subset=(state==state & measure==measure))
order.answer<-order(answer$score,answer$school)
answer1<-as.matrix(answer[order.answer,])
answer1[1,1]
}
lowest("NY","MSAT")
Incorrect answer:
[1] "UCLA"
The problem seems to be that the variables 'state' and 'measure' don't take on the values of the arguments "NY" and "MSAT" in the subset line of the function. I've experimented with '=' instead of '==' and also tried subset(data,subset=(state=="state" & measure=="measure")), but can't find a solution.

Answer 1

There seems to be something going awry in your function call related to the fact that your function arguments are the same as your data columns because this works
lowest <- function(State, Measure){
answer<-subset(data,subset=(state==State & measure==Measure))
order.answer<-order(answer$score,answer$school)
answer1<-as.matrix(answer[order.answer,])
answer1[1,1]
}
##
> lowest("NY","MSAT")
[1] "Oswego"
UPDATE: After mulling this over a bit more, I think I can offer a little more detail about what is going on internally. Notice that your first (manual) subset worked correctly above:
> subset(data,subset=(state=="NY" & measure=="MSAT"))
school state measure score
1 NYU NY MSAT 590
4 FIT NY MSAT 759
5 Oswego NY MSAT 550
However, notice that if we create two objects in the global environment, state <- "NY" and measure <- "MSAT", this does not work
state <- "NY"
measure <- "MSAT"
> subset(data,state==state & measure==measure)
school state measure score
1 NYU NY MSAT 590.0
2 BYU UT MSAT 490.0
3 USC CA GPA 2.9
4 FIT NY MSAT 759.0
5 Oswego NY MSAT 550.0
6 UCLA CA GPA 1.2
7 USF CA GPA 3.1
8 Columbia NY GPA 3.2
The reason (I believe) has to do with the R's scope resolution mechanism and how it operates within functions. When a function is called in R, this function call creates a (temporary?) environment where the objects in that function reside in the local frame, i.e. they are prioritized in the sense that if you have a variable x in the global environment, and a function defined as foo <- function(x){do something interesting to x}, that do something interesting is acting on the object passed into the foo via the argument x, and not the object x in the global environment. However, R uses dynamic scoping, meaning that if the object referred to in the local frame is not found, the interpreter will recursively search through the hierarchy of frames / environments until it finds the object referred to. So if instead of foo <- function(x){do something interesting to x} you had foo(z){do something interesting to x}, but you still had x defined in the global environment, rather than result in an error, the function would search through the call stack until it found the object x to "do something interesting" to.
In the example directly above, the reason subset(data,state==state & measure==measure) did not produce the desired results and subset(data,state=="NY & measure=="MSAT") did is because in the subset(data, ...) function call, data and all of its columns were in the local scope, i.e. the columns state and measure were prioritized over the objects state and measure in the global environment. Therefore, the subsetting condition state==state & measure==measure evaluated as TRUE for each row of data, and so the "subset" returned was the entirety of data. Now, if we do
State <- "NY"
Measure <- "MSAT"
> subset(data, state==State & measure==Measure)
school state measure score
1 NYU NY MSAT 590
4 FIT NY MSAT 759
5 Oswego NY MSAT 550
this works fine, because since State and Measure are not found in the local frame of the subset function call, the interpreter will keep searching through environments until it first encounters these objects (in this case, it finds them in the global environment). This is why when I changed the arguments to State and Measure (and made the respective changes in the function body) in your function lowest it produced the desired results - really you could change them to just about anything, as long as the names do not clash with the column names of data, but capitalizing their first letter was a quick fix.