統計學(二) 筆記 - 期末專案 國內股票型基金分析

#### file name
#merge.data.R 
####
library(magrittr)
library(quantmod)
library(agricolae)
library(tibble)
setwd("D:\\NTUT\\大二下\\統計\\報告")

data_merge <- function(folder_name){
  #"一般股票型"
  path <- file.path(getwd(), "基金CSV", folder_name)
  files <- dir(path)
  data <- tibble(YM = character(), name = character(), return = numeric(),
                 sharpe = numeric(), sd = numeric())
  
  for (fname in files){
    
    path <- file.path(getwd(), "基金CSV", folder_name, fname)
    fund <- read.csv(path)
    fname <- gsub(".csv", "", fname)
    fund$date <- fund$日期 %>% as.Date() %>% as.character() #將日期與 yahoo 股價一致
    fund$漲跌幅 <- gsub("%", "", fund$漲跌幅) %>% as.numeric()  #子串轉數字
    fund <- fund[-grep("2021", fund$date) ,] #消除 2021 的資料
    fund <- fund[-grep("2016", fund$date) ,] #消除 2016 的資料
    if(nrow(fund) <= 0) next
    
    fund$YM <- substr(fund$date, 1, 4)#year with month 取出年份與月份
    YMs <- unique(fund$YM) 

    #將基金的每日資料合併成每年的資料
    for(ym in YMs){
      sub_fund <- fund[fund$YM == ym, ]
      return <- (tail(sub_fund$淨值,1) - head(sub_fund$淨值,1)) / head(sub_fund$淨值,1) #計算報酬率
      data <- data %>% add_row(YM = ym, name = fname,
                         return = return,
                         sharpe = (mean(sub_fund$漲跌幅) - 0.0000232) / sd(sub_fund$漲跌幅) * (sqrt(252)), # sharpe ratio
                         sd = sd(sub_fund$漲跌幅)) # 標準差
    #sharpe ratio 的無風險利率使用台灣銀行一天的利息做標準 0.0000232 
    }
  }
  return(data)
}

#### file name
#fund.R
####
library(magrittr)
library(quantmod)
library(agricolae)
library(stats)
setwd("D:\\NTUT\\大二下\\統計\\報告")

source("./merge_data.R", encoding = "UTF-8")
source("./loading_data.R", encoding = "UTF-8")

fund_type <- c("一般股票型", "中小型", "科技型")
#data_return <- function(folder_name)
data <- data_merge("一般股票型")
#write.csv(data, "一般股票型.csv")
write.csv(data, "./ppt材料/一般股票型統計.csv")

#報酬率
fit <- aov(formula = return ~ name, data = data) # anova 分析
anova_table <- anova(fit)
lsd <- LSD.test(fit, "name", p.adj ="none", alpha = 0.05) #判斷 LSD 分析
lsd$groups$name <- lsd$groups %>% row.names()
row.names(lsd$groups) <- NULL #將 rownames 轉為 index
View(lsd$groups)
good_funds <- lsd$groups$name[grep("a", lsd$groups$groups)]
data <- data[data$name %in% good_funds, ]

#sharpe
fit <- aov(formula = sharpe ~ name, data = data) #anova 分析
anova_table <- anova(fit)
lsd <- LSD.test(fit, "name", p.adj ="none", alpha = 0.05) #判斷 LSD 分析
lsd$groups$name <- lsd$groups %>% row.names()
row.names(lsd$groups) <- NULL #將 rownames 轉為 index
View(lsd$groups)
good_funds <- lsd$groups$name[grep("a", lsd$groups$groups)]
data <- data[data$name %in% good_funds, ]

#### file name
#var.R
####
library(magrittr)
library(quantmod)
library(agricolae)
library(stats)
setwd("D:\\NTUT\\大二下\\統計\\報告")

source("./merge_data.R", encoding = "UTF-8")

#在使用此函式前，必須先使用 *_fund.R 先判斷哪些基金最好

fund_type <- c("一般股票型", "中小型", "科技型")
#data_return <- function(folder_name)
normal_fund <- data_merge("一般股票型")
small_fund <- data_merge("中小型")
tech_fund <- data_merge("科技型")

write.csv(normal_fund[normal_fund$name == "安聯台灣智慧基金",], "./ppt材料/安聯台灣智慧基金統計.csv")
write.csv(small_fund[small_fund$name == "復華中小精選基金",], "./ppt材料/復華中小精選基金統計.csv")
write.csv(tech_fund[tech_fund$name == "安聯台灣科技基金",], "./ppt材料/安聯台灣科技基金統計.csv")
normal_fund <- normal_fund$return[normal_fund$name == "安聯台灣智慧基金"] 
small_fund <- small_fund$return[small_fund$name == "復華中小精選基金"]
tech_fund <- tech_fund$return[tech_fund$name == "安聯台灣科技基金"]


var(normal_fund)
var(small_fund)
var(tech_fund)
#Reject
var.test(normal_fund, 
         small_fund,
         alternative = "two.sided") #less 表示 A 股票比 B 股票標準差更大

#### file name
# lm.R 
####
library(magrittr)
library(quantmod)
library(agricolae)
library(tibble)
library(ggfortify)
source("./merge_data.R", encoding = "UTF-8")

setwd("D:\\NTUT\\大二下\\統計\\報告")

fund_type <- c("一般股票型", "中小型", "科技型")
#data_return <- function(folder_name)

path <- file.path(getwd(), "基金CSV", "科技型", "安聯台灣科技基金.csv")

fund <- read.csv(path)
fund$date <- fund$日期 %>% as.Date() %>% as.character()

fund$return <- gsub("%", "", fund$漲跌幅) %>% as.numeric() 
fund$return <- fund$return * 0.01
date <- fund$date

#基金的前十大股票
codes <- c("2330.TW", "3529.TWO", "2492.TW", "2327.TW", "2337.TW", "8016.TW", "6510.TWO",
           "6732.TWO", "4966.TWO", "3035.TW")
for (code in codes){ #找出所有股票的共同日期
  stock <- getSymbols(code, auto.assign = FALSE , from = "2016-01-01" ) %>% na.omit() %>% data.frame()
  names(stock) <- c("open", "high", "low", "close", "vol", "adjPrice")
  stock$date <- rownames(stock) %>% as.Date() %>% as.character()
  rownames(stock) <- NULL
  date <- intersect(date, stock$date)
}

data <- tibble(date = date, return = fund$return[fund$date %in% date])
for(code in codes){ #找出所有股票的共同日期的報酬率
  stock <- getSymbols(code, auto.assign = FALSE , from = "2016-01-01" ) %>% na.omit() %>% data.frame()
  names(stock) <- c("open", "high", "low", "close", "vol", "adjPrice")
  stock$date <- rownames(stock) %>% as.Date() %>% as.character()
  stock$last_close <- c(Inf, head(stock$close, length(stock$close)-1))
  stock$return <- (stock$close - stock$last_close) / stock$last_close
  rownames(stock) <- NULL
  data[ , code] <- stock$return[stock$date %in% date]
}


#共線性測試，
cor.test <- cor(data[3:ncol(data)])
cor.test

#將共線性過高的刪除
data["2327.TW"] <- NULL
write.csv(data, "科技股回歸資料.csv")

#迴歸分析
model <- lm(formula = return ~ ., data = data[, 2:ncol(data)]) 
View(model)
sum_model <- summary(model)
View(sum_model)

#殘差圖、標準常態圖
autoplot(model)