jhu_prgwr_w4.Rmd

---
title: "JHU - Simulation & Profiling"
author: 郭耀仁
date: "`r Sys.Date()`"
output: 
  revealjs::revealjs_presentation:
    theme: black
    highlight: zenburn
    center: true
---

## 作業來源

[Coursera](https://www.coursera.org/) 的 [Programming with R](https://www.coursera.org/learn/r-programming) 課程第四週作業

## 資料集

[Assignment3-data](https://storage.googleapis.com/jhu_coursera_data/Assignment3-data.zip)

## 資料內容

- outcome-of-care-measures.csv 30 天心臟病就診資料
- hospital-data.csv 4,000+ 醫院的資訊
- Hospital_Revised_Flatfiles.pdf 編碼手冊

# 練習一

## 練習一說明

- 將 outcome-of-care-measures.csv 讀入，並且觀察第 11 個變數的分佈

```{r}
library(ggplot2)

outcome <- read.csv("~/Downloads/Assignment3-data/outcome-of-care-measures.csv", stringsAsFactors = FALSE)
outcome[, 11] <- as.numeric(outcome[, 11])
```

## 練習一範例

```{r echo=FALSE, warning=FALSE}
ggplot(outcome, mapping = aes(x = Hospital.30.Day.Death..Mortality..Rates.from.Heart.Attack)) +
  geom_histogram(bins = 40, fill = rgb(0.8, 0, 0, 0.5)) +
  theme_minimal() +
  ggtitle("30-day mortality rates for heart attack") +
  xlab("30-day mortality rates")
```

# 練習二

## 練習二函數結構

`best(state, outcome)`

找出指定州別裡面最好的醫院

## 練習二函數提示

- `outcome` 參數可以輸入 "heart attack"、"heart failure" 或 "pneumonia"
- 如果有兩個以上醫院的 outcome 相同，則依照字母順序排序後回傳
- `suppressWarnings()` 可以消除警告訊息
- 假如輸入無效的 `state` 名稱，使用 `stop()` 函數回傳訊息 invalid state
- 假如輸入無效的 `outcome` 參數，使用 `stop()` 函數回傳訊息 invalid outcome

## 練習二函數範例

```{r echo=FALSE}
best <- function(state, outcome) {
  outcome_df <- read.csv("~/Downloads/Assignment3-data/outcome-of-care-measures.csv", stringsAsFactors = FALSE)
  unique_state_names <- unique(outcome_df$State)
  unique_outcomes <- c("heart attack", "heart failure", "pneumonia")
  cols_to_numeric <- c("Hospital.30.Day.Death..Mortality..Rates.from.Heart.Attack", "Hospital.30.Day.Death..Mortality..Rates.from.Heart.Failure", "Hospital.30.Day.Death..Mortality..Rates.from.Pneumonia")
  for (char_col in cols_to_numeric) {
    suppressWarnings(
      outcome_df[, char_col] <- as.numeric(outcome_df[, char_col])
    )
  }
  if (!(state %in% unique_state_names)) {
    stop("invalid state")
  } else if (!(outcome %in% unique_outcomes)) {
    stop("invalid outcome")
  } else {
    state_outcome <- outcome_df[outcome_df$State == state, ]
    if (outcome == "heart attack") {
      lowest_indice <- which(state_outcome$Hospital.30.Day.Death..Mortality..Rates.from.Heart.Attack == min(state_outcome$Hospital.30.Day.Death..Mortality..Rates.from.Heart.Attack, na.rm = TRUE))
      hospital_names <- state_outcome$Hospital.Name[lowest_indice]
      hospital_names <- sort(hospital_names)
      return(hospital_names)
    } else if (outcome == "heart failure") {
      lowest_indice <- which(state_outcome$Hospital.30.Day.Death..Mortality..Rates.from.Heart.Failure == min(state_outcome$Hospital.30.Day.Death..Mortality..Rates.from.Heart.Failure, na.rm = TRUE))
      hospital_names <- state_outcome$Hospital.Name[lowest_indice]
      hospital_names <- sort(hospital_names)
      return(hospital_names)
    } else if (outcome == "pneumonia") {
      lowest_indice <- which(state_outcome$Hospital.30.Day.Death..Mortality..Rates.from.Pneumonia == min(state_outcome$Hospital.30.Day.Death..Mortality..Rates.from.Pneumonia, na.rm = TRUE))
      hospital_names <- state_outcome$Hospital.Name[lowest_indice]
      hospital_names <- sort(hospital_names)
      return(hospital_names)
    }
  }
}
```

```{r}
best("TX", "heart attack")
best("TX", "heart failure")
best("MD", "heart attack")
```

----

```{r error=TRUE}
best("MD", "pneumonia")
best("BB", "heart attack")
best("NY", "hert attack")
```

# 練習三

## 練習三函數結構

`rankhospital(state, outcome, num = "best")`

將一個州依照 `outcome` 參數選出最好、最差或特定排名的醫院

## 練習三函數提示

- `outcome` 參數可以輸入 "heart attack"、"heart failure" 或 "pneumonia"
- 如果有兩個以上醫院的 outcome 相同，則依照字母順序排序後回傳
- 假如輸入無效的 `state` 名稱，使用 `stop()` 函數回傳訊息 invalid state
- 假如輸入無效的 `outcome` 參數，使用 `stop()` 函數回傳訊息 invalid outcome

## 練習三函數範例

```{r echo = FALSE}
rankhospital <- function(state, outcome, num = "best") {
  outcome_df <- read.csv("~/Downloads/Assignment3-data/outcome-of-care-measures.csv", stringsAsFactors = FALSE)
  unique_state_names <- unique(outcome_df$State)
  unique_outcomes <- c("heart attack", "heart failure", "pneumonia")
  cols_to_numeric <- c("Hospital.30.Day.Death..Mortality..Rates.from.Heart.Attack", "Hospital.30.Day.Death..Mortality..Rates.from.Heart.Failure", "Hospital.30.Day.Death..Mortality..Rates.from.Pneumonia")
  for (char_col in cols_to_numeric) {
    suppressWarnings(
      outcome_df[, char_col] <- as.numeric(outcome_df[, char_col])
    )
  }
  if (!(state %in% unique_state_names)) {
    stop("invalid state")
  } else if (!(outcome %in% unique_outcomes)) {
    stop("invalid outcome")
  } else {
    state_outcome <- outcome_df[outcome_df$State == state, ]
    if (outcome == "heart attack") {
      if (num == "best") {
        lowest_indice <- which(state_outcome$Hospital.30.Day.Death..Mortality..Rates.from.Heart.Attack == min(state_outcome$Hospital.30.Day.Death..Mortality..Rates.from.Heart.Attack, na.rm = TRUE))
        hospital_names <- state_outcome$Hospital.Name[lowest_indice]
        hospital_names <- sort(hospital_names)
        return(hospital_names)
      } else if (num == "worst") {
        lowest_indice <- which(state_outcome$Hospital.30.Day.Death..Mortality..Rates.from.Heart.Attack == max(state_outcome$Hospital.30.Day.Death..Mortality..Rates.from.Heart.Attack, na.rm = TRUE))
        hospital_names <- state_outcome$Hospital.Name[lowest_indice]
        hospital_names <- sort(hospital_names)
        return(hospital_names)
      } else if (num > nrow(state_outcome)) {
        return(NA)
      } else if (is.numeric(num)) {
        state_outcome_ordered <- state_outcome[order(state_outcome$Hospital.30.Day.Death..Mortality..Rates.from.Heart.Attack, state_outcome$Hospital.Name), ]
        return(state_outcome_ordered$Hospital.Name[num])
      } else {
        stop("invalid num")
      }
    } else if (outcome == "heart failure") {
      if (num == "best") {
        lowest_indice <- which(state_outcome$Hospital.30.Day.Death..Mortality..Rates.from.Heart.Failure == min(state_outcome$Hospital.30.Day.Death..Mortality..Rates.from.Heart.Failure, na.rm = TRUE))
        hospital_names <- state_outcome$Hospital.Name[lowest_indice]
        hospital_names <- sort(hospital_names)
        return(hospital_names)
      } else if (num == "worst") {
        lowest_indice <- which(state_outcome$Hospital.30.Day.Death..Mortality..Rates.from.Heart.Failure == max(state_outcome$Hospital.30.Day.Death..Mortality..Rates.from.Heart.Failure, na.rm = TRUE))
        hospital_names <- state_outcome$Hospital.Name[lowest_indice]
        hospital_names <- sort(hospital_names)
        return(hospital_names)
      } else if (num > nrow(state_outcome)) {
        return(NA)
      } else if (is.numeric(num)) {
        state_outcome_ordered <- state_outcome[order(state_outcome$Hospital.30.Day.Death..Mortality..Rates.from.Heart.Failure, state_outcome$Hospital.Name), ]
        return(state_outcome_ordered$Hospital.Name[num])
      } else {
        stop("invalid num")
      }
    } else if (outcome == "pneumonia") {
      if (num == "best") {
        lowest_indice <- which(state_outcome$Hospital.30.Day.Death..Mortality..Rates.from.Pneumonia == min(state_outcome$Hospital.30.Day.Death..Mortality..Rates.from.Pneumonia, na.rm = TRUE))
        hospital_names <- state_outcome$Hospital.Name[lowest_indice]
        hospital_names <- sort(hospital_names)
        return(hospital_names)
      } else if (num == "worst") {
        lowest_indice <- which(state_outcome$Hospital.30.Day.Death..Mortality..Rates.from.Pneumonia == max(state_outcome$Hospital.30.Day.Death..Mortality..Rates.from.Pneumonia, na.rm = TRUE))
        hospital_names <- state_outcome$Hospital.Name[lowest_indice]
        hospital_names <- sort(hospital_names)
        return(hospital_names)
      } else if (num > nrow(state_outcome)) {
        return(NA)
      } else if (is.numeric(num)) {
        state_outcome_ordered <- state_outcome[order(state_outcome$Hospital.30.Day.Death..Mortality..Rates.from.Pneumonia, state_outcome$Hospital.Name), ]
        return(state_outcome_ordered$Hospital.Name[num])
      } else {
        stop("invalid num")
      }
    }
  }
}
```

```{r}
rankhospital("TX", "heart failure", 4)
rankhospital("MD", "heart attack", "worst")
rankhospital("MN", "heart attack", 5000)
```

# 練習四

## 練習四函數結構

`rankall(outcome, num = "best")`

將全國的醫院依照 `outcome` 參數選出各州最好、最差或特定排名的醫院

## 練習四函數提示

- `outcome` 參數可以輸入 "heart attack"、"heart failure" 或 "pneumonia"
- 如果有兩個以上醫院的 outcome 相同，則依照字母順序排序後回傳
- 假如輸入無效的 `outcome` 參數，使用 `stop()` 函數回傳訊息 invalid outcome

## 練習四函數範例

```{r echo = FALSE}
rankall <- function(outcome, num = "best") {
  outcome_df <- read.csv("~/Downloads/Assignment3-data/outcome-of-care-measures.csv", stringsAsFactors = FALSE)
  unique_state_names <- sort(unique(outcome_df$State))
  unique_outcomes <- c("heart attack", "heart failure", "pneumonia")
  cols_to_numeric <- c("Hospital.30.Day.Death..Mortality..Rates.from.Heart.Attack", "Hospital.30.Day.Death..Mortality..Rates.from.Heart.Failure", "Hospital.30.Day.Death..Mortality..Rates.from.Pneumonia")
  for (char_col in cols_to_numeric) {
    suppressWarnings(
      outcome_df[, char_col] <- as.numeric(outcome_df[, char_col])
    )
  }
  if (!(outcome %in% unique_outcomes)) {
    stop("invalid outcome")
  } else {
    if (outcome == "heart attack") {
      if (num == "best") {
        result_df <- data.frame(best_hospitals = c(), states = c())
        for (state in unique_state_names) {
          each_state <- outcome_df[outcome_df$State == state, ]
          each_state_ordered <- each_state[order(each_state$Hospital.30.Day.Death..Mortality..Rates.from.Heart.Attack, each_state$Hospital.Name), ]
          best_in_each_state <- each_state_ordered$Hospital.Name[1]
          df_to_bind <- data.frame(best_hospitals = best_in_each_state, states = state)
          result_df <- rbind(result_df, df_to_bind)
        }
        names(result_df) <- c("hospital", "state")
        return(result_df)
      } else if (num == "worst") {
        result_df <- data.frame(worst_hospitals = c(), states = c())
        for (state in unique_state_names) {
          each_state <- outcome_df[outcome_df$State == state, ]
          each_state_ordered <- each_state[order(each_state$Hospital.30.Day.Death..Mortality..Rates.from.Heart.Attack, each_state$Hospital.Name, decreasing = TRUE), ]
          worst_in_each_state <- each_state_ordered$Hospital.Name[length(each_state_ordered$Hospital.Name)]
          df_to_bind <- data.frame(worst_hospitals = worst_in_each_state, states = state)
          result_df <- rbind(result_df, df_to_bind)
        }
        names(result_df) <- c("hospital", "state")
        return(result_df)
      } else if (is.numeric(num)) {
        result_df <- data.frame(best_hospitals = c(), states = c())
        for (state in unique_state_names) {
          each_state <- outcome_df[outcome_df$State == state, ]
          ordered_in_each_state <- each_state[order(each_state$Hospital.30.Day.Death..Mortality..Rates.from.Heart.Attack, each_state$Hospital.Name), ]
          if (num > nrow(ordered_in_each_state)) {
            df_to_bind <- data.frame(best_hospitals = NA, states = state)
            result_df <- rbind(result_df, df_to_bind)
          } else {
            df_to_bind <- data.frame(best_hospitals = ordered_in_each_state$Hospital.Name[num], states = state)
            result_df <- rbind(result_df, df_to_bind)
          }
        }
        names(result_df) <- c("hospital", "state")
        return(result_df)
      }
    } else if (outcome == "heart failure") {
      if (num == "best") {
        result_df <- data.frame(best_hospitals = c(), states = c())
        for (state in unique_state_names) {
          each_state <- outcome_df[outcome_df$State == state, ]
          each_state_ordered <- each_state[order(each_state$Hospital.30.Day.Death..Mortality..Rates.from.Heart.Failure, each_state$Hospital.Name), ]
          best_in_each_state <- each_state_ordered$Hospital.Name[1]
          df_to_bind <- data.frame(best_hospitals = best_in_each_state, states = state)
          result_df <- rbind(result_df, df_to_bind)
        }
        names(result_df) <- c("hospital", "state")
        return(result_df)
      } else if (num == "worst") {
        result_df <- data.frame(worst_hospitals = c(), states = c())
        for (state in unique_state_names) {
          each_state <- outcome_df[outcome_df$State == state, ]
          each_state_ordered <- each_state[order(each_state$Hospital.30.Day.Death..Mortality..Rates.from.Heart.Failure, each_state$Hospital.Name, decreasing = TRUE), ]
          worst_in_each_state <- each_state_ordered$Hospital.Name[length(each_state_ordered$Hospital.Name)]
          df_to_bind <- data.frame(worst_hospitals = worst_in_each_state, states = state)
          result_df <- rbind(result_df, df_to_bind)
        }
        names(result_df) <- c("hospital", "state")
        return(result_df)
      } else if (is.numeric(num)) {
        result_df <- data.frame(best_hospitals = c(), states = c())
        for (state in unique_state_names) {
          each_state <- outcome_df[outcome_df$State == state, ]
          ordered_in_each_state <- each_state$Hospital.Name[order(each_state$Hospital.30.Day.Death..Mortality..Rates.from.Heart.Failure, each_state$Hospital.Name), ]
          if (num > length(ordered_in_each_state)) {
            df_to_bind <- data.frame(best_hospitals = NA, states = state)
            result_df <- rbind(result_df, df_to_bind)
          } else {
            df_to_bind <- data.frame(best_hospitals = ordered_in_each_state[num], states = state)
            result_df <- rbind(result_df, df_to_bind)
          }
        }
        names(result_df) <- c("hospital", "state")
        return(result_df)
      }
    } else if (outcome == "pneumonia") {
      if (num == "best") {
        result_df <- data.frame(best_hospitals = c(), states = c())
        for (state in unique_state_names) {
          each_state <- outcome_df[outcome_df$State == state, ]
          each_state_ordered <- each_state[order(each_state$Hospital.30.Day.Death..Mortality..Rates.from.Pneumonia, each_state$Hospital.Name), ]
          best_in_each_state <- each_state_ordered$Hospital.Name[1]
          df_to_bind <- data.frame(best_hospitals = best_in_each_state, states = state)
          result_df <- rbind(result_df, df_to_bind)
        }
        names(result_df) <- c("hospital", "state")
        return(result_df)
      } else if (num == "worst") {
        result_df <- data.frame(worst_hospitals = c(), states = c())
        for (state in unique_state_names) {
          each_state <- outcome_df[outcome_df$State == state, ]
          each_state_ordered <- each_state[order(each_state$Hospital.30.Day.Death..Mortality..Rates.from.Pneumonia, each_state$Hospital.Name, decreasing = TRUE), ]
          worst_in_each_state <- each_state_ordered$Hospital.Name[1]
          df_to_bind <- data.frame(worst_hospitals = worst_in_each_state, states = state)
          result_df <- rbind(result_df, df_to_bind)
        }
        names(result_df) <- c("hospital", "state")
        return(result_df)
      } else if (is.numeric(num)) {
        best_hospitals <- c()
        states <- c()
        result_df <- data.frame(best_hospitals, states)
        for (state in unique_state_names) {
          each_state <- outcome_df[outcome_df$State == state, ]
          ordered_in_each_state <- each_state$Hospital.Name[order(each_state$Hospital.30.Day.Death..Mortality..Rates.from.Pneumonia, each_state$Hospital.Name), ]
          if (num > length(ordered_in_each_state)) {
            df_to_bind <- data.frame(best_hospitals = NA, states = state)
            result_df <- rbind(result_df, df_to_bind)
          } else {
            df_to_bind <- data.frame(best_hospitals = ordered_in_each_state[num], states = state)
            result_df <- rbind(result_df, df_to_bind)
          }
        }
        names(result_df) <- c("hospital", "state")
        return(result_df)
      }
    }
  }
}
```

```{r}
head(rankall("heart attack", 20), 10)
tail(rankall("pneumonia", "worst"), 3)
```

----

```{r}
tail(rankall("heart failure"), 10)
```