rankall.R

rankall<- function(outcome, num = "best") {
        
        # check the validity of the outcome and distribute the according colunm to the outcome
        if (outcome == "heart attack") # heart attack
                colnum <- 11
        else if (outcome == "heart failure") # heart failure
                colnum <- 17
        else if (outcome == "pneumonia")  # pneumonia
                colnum <- 23
        else 
                stop("invalid outcome")
        
        
        # check the validity of the num
        if (!is.numeric(num) && !(num %in% c("best", "worst")))
                stop("invalid num")
        
        data <- read.csv("rprog-data-ProgAssignment3-data/outcome-of-care-measures.csv"
                         , colClasses = "character")
        
        # combine the 3 useful columns together and remove na
        fd   <- as.data.frame(cbind(data[, 2],  # hospital
                                    data[, 7],  # state
                                    data[, colnum]), # outcome
                              stringsAsFactors = FALSE) 
        colnames(fd) <- c("hospital", "state", outcome)
        fd[[outcome]] <- as.numeric(fd[[outcome]])
        fd <- na.omit(fd)
        
        if (num == "best")
                num <- 1
        
        # for worst, we will reverse the order of rank, and take the 1st one
        reverse_sort <- FALSE
        if (num == "worst") {
                reverse_sort <- TRUE
                num <- 1
        }
        
        by_state <- split(fd, fd$state)
        
        result <- data.frame()
        
        for (i in seq_along(by_state)) {
                r <- by_state[[i]][order(by_state[[i]][outcome], 
                                         by_state[[i]]["hospital"], 
                                         decreasing = reverse_sort)
                                   , ]
                r <- data.frame(list(hospital = r[num, "hospital"], state = r[1, "state"]))
                result <- rbind(result, r)
        }
        result
}