r4ds

迷茫的日子里也不能忘了要學(xué)習(xí)新知識(shí)技能以及總結(jié)歸納啊鲜戒。最近開(kāi)始學(xué)習(xí)《R數(shù)據(jù)科學(xué)》中文版，這確實(shí)是本好書(shū)抹凳，有些知識(shí)點(diǎn)的層次關(guān)系及解釋讓我明白了許多之前對(duì)R一知半解的地方遏餐。以接下來(lái)的一系列筆記來(lái)歸納建立自己的知識(shí)體系，并一起分享交流學(xué)習(xí)赢底。

第一章： 使用ggplot2進(jìn)行數(shù)據(jù)可視化

1. ggplot()失都，在此函數(shù)種添加的映射會(huì)作為全局變量應(yīng)用到圖中的每個(gè)幾何對(duì)象種。

2. 圖層geom_point點(diǎn)圖層

3. 映射數(shù)據(jù)為圖形屬性 mapping=aes()幸冻，要想將圖形屬性映射為變量粹庞，需要在函數(shù)aes()中將圖形屬性的名稱(chēng)和變量的名稱(chēng)關(guān)聯(lián)起來(lái)。

4. 標(biāo)度變化：將變量（數(shù)據(jù)）分配唯一的圖形屬性水平嘁扼。

5. 手動(dòng)設(shè)置圖形屬性信粮，此是在geom_point()層面黔攒。此時(shí)趁啸，這個(gè)顏色是不會(huì)傳達(dá)變量數(shù)據(jù)的信息。

6. 分層facet：facet_grid()可以通過(guò)兩個(gè)變量對(duì)圖分層`facet_grid(drv_cyl)或(.cyl)

7. 分組aes(group)此種按照?qǐng)D形屬性的分組不用添加圖例督惰，也不用為幾何對(duì)象添加區(qū)分特征

8. 統(tǒng)計(jì)變換：繪圖時(shí)用來(lái)計(jì)算新數(shù)據(jù)的算法稱(chēng)為stat(statistical transformation,統(tǒng)計(jì)變化)不傅。比如對(duì)于geom_bar()默認(rèn)是只對(duì)一個(gè)數(shù)據(jù)x映射，其統(tǒng)計(jì)變化后生成數(shù)據(jù)x種的每個(gè)值的count數(shù)赏胚。

   • 每個(gè)幾何對(duì)象函數(shù)都有一個(gè)默認(rèn)的統(tǒng)計(jì)變換访娶，每個(gè)統(tǒng)計(jì)變換函數(shù)都有一個(gè)默認(rèn)的幾何對(duì)象。
   • 如需要展示二維數(shù)據(jù)觉阅，geom_bar(mapping=aes(x=a,y=b),stat="identity ")

9. 圖形屬性/位置調(diào)整：

   • color,fill
   • 位置調(diào)整參數(shù)position有三個(gè)選項(xiàng)："identity","fill","dodge"
   • position="dodge"參數(shù)可分組顯示數(shù)據(jù)崖疤，將每組種的條形依次并列放置，可以輕松比較每個(gè)條形表示的具體數(shù)值典勇。
   • 數(shù)據(jù)的聚集模式無(wú)法很好確定劫哼，因?yàn)榇嬖跀?shù)據(jù)的過(guò)繪制問(wèn)題（很多彼此十分僅的點(diǎn)重疊了）position="jitter"對(duì)于geom_position()函數(shù)來(lái)說(shuō)，jitter的位置方式為抖動(dòng)會(huì)排除過(guò)繪制問(wèn)題

10. 坐標(biāo)系：

    • coord_flip()函數(shù)可以交換x軸和y軸
    • labs()：modify axis, legend, and plot labels.

mpg
str(mpg)
data<- mpg
?mpg ##查看mpg數(shù)據(jù)的說(shuō)明
ggplot(data = mpg)+geom_point(aes(x=displ,y=hwy))

ggplot(mpg)+geom_point(mapping = aes(x=displ,y=hwy,color=class),color="#EF5C4E",shape=19)
ggplot(mpg)+geom_point(mapping = aes(x=displ,y=hwy),color="#EF5C4E",shape=19)
ggplot(mpg)+geom_point(mapping = aes(x=displ,y=hwy,stroke=displ),shape=19)
## 添加兩個(gè)圖層：geom_point,geom_smooth()
ggplot(mpg)+geom_point(mapping = aes(x=displ,y=hwy,color=drv))+geom_smooth(mapping = aes(x=displ,y=hwy,linetype=drv,color=drv))
# 添加分組
ggplot(data = mpg)+geom_smooth(mapping = aes(x=displ,y=hwy,group=drv))
ggplot(data = mpg)+geom_smooth(mapping = aes(x=displ,y=hwy,color=drv),show.legend = F) ## 圖例 show.legend=F

## 在不同的圖層中添加指定不同的數(shù)據(jù)
## data=filter(mpg,class=="suv"), se=F割笙，表示去除f波動(dòng)的范圍权烧。
ggplot(data = mpg,mapping = aes(x=displ,y=hwy))+geom_point(mapping = aes(color=class))+geom_smooth(data = filter(.data = mpg,class=="suv"))

##exercices
ggplot(data = mpg,mapping = aes(x=displ,y=hwy))+geom_point()+geom_smooth(se = F)
ggplot(data = mpg,mapping = aes(x=displ,y=hwy))+geom_point()+geom_smooth(se = F,mapping = aes(group=drv))
ggplot(data = mpg,mapping = aes(x=displ,y=hwy,color=drv))+geom_point()+geom_smooth(se = F)
ggplot(data = mpg,mapping = aes(x=displ,y=hwy))+geom_point(mapping = aes(color=drv))+geom_smooth(se = F)
ggplot(data = mpg,mapping = aes(x=displ,y=hwy))+geom_point(mapping = aes(color=drv))+geom_smooth(mapping = aes(linetype=drv),se = F)
ggplot(data = mpg,mapping = aes(x=displ,y=hwy))+geom_point(mapping = aes(color=drv))

### 統(tǒng)計(jì)變換
ggplot(data = mpg,mapping = aes(x=displ,y=hwy))+geom_point()+geom_smooth(se = F)
ggplot(data=diamonds)+stat_summary(mapping = aes(x=cut,y=depth))
ggplot(data=diamonds)+geom_boxplot(mapping = aes(x=cut,y=price))
ggplot(data=diamonds)+geom_bar(mapping = aes(x=cut))
ggplot(data=diamonds)+geom_bar(mapping = aes(x=cut,y=..prop..),group=2)

### 圖形調(diào)整,位置調(diào)整
ggplot(diamonds)+geom_bar(mapping = aes(x=cut,fill=cut),color="black")+scale_fill_brewer(palette = "Set3")
ggplot(diamonds)+geom_bar(mapping = aes(x=cut,fill=clarity))+scale_fill_brewer(palette = "Set2")
ggplot(diamonds)+geom_bar(mapping=aes(x=cut,color=clarity),position = "dodge")+scale_fill_brewer(palette = "Set2")
ggplot(mpg)+geom_point(mapping = aes(x=displ,y=hwy),position = "jitter")
##exercises
ggplot(mpg,mapping = aes(x=cty,y=hwy))+geom_point(position = "jitter")+geom_smooth(color="black")
ggplot(mpg,mapping = aes(x=cty,y=hwy))+geom_jitter()
ggplot(mpg,mapping = aes(x=cty,y=hwy))+geom_count()
ggplot(mpg)+geom_boxplot(mapping = aes(x=manufacturer,y=hwy),position = "identity")
?geom_boxplot

###1.9 坐標(biāo)系
ggplot(mpg,mapping = aes(x=class,y=hwy))+geom_boxplot()+coord_flip()
nz <- map_data("nz")
?map_data
ggplot(data=diamonds)+geom_bar(mapping = aes(x=cut,fill=cut),show.legend = FALSE)+theme(aspect.ratio = 1)+labs()
bar+scale_color_brewer(palette = "Set2")
bar+coord_flip()
bar+coord_polar()

第二章：工作流：基礎(chǔ) Workflow：basics

1. 賦值：小技巧，alt+減號(hào)會(huì)自動(dòng)輸入賦值符號(hào)<- 并在前后加空格
2. 對(duì)象：用snake_case命名法小寫(xiě)字母伤溉，以_分割般码。
3. Rstudio中快捷查看命令：Alt+Shift+K

第三章：使用dplyr進(jìn)行數(shù)據(jù)轉(zhuǎn)換

1. 特殊的data.frametibble。
2. 變量variable類(lèi)型：
   • int：
   • dbl(num的一種乱顾？):雙精度浮點(diǎn)數(shù)變量板祝，或稱(chēng)實(shí)數(shù)。
   • chr：字符向量/字符串
   • dttm：日期+時(shí)間
   • lgl：邏輯型變量
   • fctr（factor）：因子
   • date：日期型變量
3. 基礎(chǔ)函數(shù)：filter(),arrange(),select(),mutate(),summarize(),group_by()
4. 使用filter()篩選:
   • filter(flights, arr_delay<=10).
   • 比較運(yùn)算符 ==, !=, <=
   • 邏輯運(yùn)算符 x & !y, x|y, xor(x,y)
   • 缺失值 NA, is.na()

## filter()
(jan1 <- filter(flights,month==1,day==1))
(dec25 <- filter(flights,month==12,day==25))
filter(flights,month>=11)
(nov_dec <- filter(flights,month %in% c(11,12)))
filter(flights,!(arr_delay<=120 | dep_delay<=120))
NA>=10
x <- NA
is.na(x)
df <- tibble(x=c(1,NA,2))
filter(df,x>1)
filter(df,is.na(x)|x>1)
### exercise
filter(flights,carrier %in% c("UA","AA","DL"))
filter(flights,month %in% c(7,8,9))
filter(flights,arr_delay>120 & dep_delay==0)
filter(flights,dep_delay >= 60 & (dep_delay-arr_delay>=30)) 
filter(flights,dep_time ==2400| dep_time<=600)
filter(flights,is.na(dep_time))

5. 使用 arrange()按照列(variable)的值進(jìn)行排序
   • desc倒序
   • 缺失值排在最后走净，若想提前可desc(is.na())

# arrange()
arrange(flights,desc(dep_delay,arr_delay)) #降序排列
arrange(flights,desc(is.na(dep_delay),arr_delay)) ##將NA值排列到前面

6. 使用select()選擇列：（數(shù)據(jù)集會(huì)有成千上萬(wàn)個(gè)變量券时，select選出變量的子集）
   • 選出year~day之間的列：select(flights, year:day)
   • 選出排除year~day列：select(flights,-(year:dat))
   • 匹配變量中的名稱(chēng)：starts_with(""), ends_with(), matches("")
   • rename()對(duì)列重新命名
   • everything()輔助函數(shù)來(lái)將某幾列提前囊嘉。

## select()
select(flights,year:day)
select(flights,-(year:day)) ## 不包括year:day
select(flights,starts_with("s"))
select(flights,ends_with("e"))
select(flights,matches("time"))
select(flights,matches("(.)\\1"))

rename(flights,tail_num=tailnum) ##對(duì)變量進(jìn)行重命名
select(flights,-(month:day),everything()) ## 結(jié)合everything()輔助函數(shù) 對(duì)某幾列提前, 置后同理
select(flights, hour:time_hour,everything())
###exercise
select(flights,year,year,year)
select(flights,one_of(c("year","month","day","dep_delay")))
select(flights,contains("TIME"))

7. 使用mutate()添加新的列/變量：
   • mutate() 新列添加到已有列的后面；
   • transmute 只保留新的變量革为。
   • 常用的運(yùn)算符號(hào)：求整%/%扭粱，求余%%，偏移函數(shù)lead(), lag()震檩，累加和和滾動(dòng)聚合琢蛤，邏輯比較，排秩抛虏。

# mutate() 在tibble后 添加新變量/列
flights_sml <- select(flights,year:day,matches("_delay"),distance,air_time)
flights_sml
mutate(flights_sml,flying_delay=arr_delay-dep_delay,speed=distance/air_time * 60 )
flights_sml
transmute(flights,gain=arr_delay-dep_delay,hour=air_time/60,gain_per_hour=gain/hour)

mutate(flights,dep_time=((dep_time%/%100 * 60)+(dep_time%%100))) ## 會(huì)直接在flights中改動(dòng)dep_time
flights
transmute(flights,air_time,duration=(arr_time-dep_time),arr_delay)
1:3+1:10
1:10+1:3
1:10
?cos

8. 使用summarize()進(jìn)行分組摘要：
   • 與group_by一起使用博其，將整個(gè)數(shù)據(jù)集的單位縮小為單個(gè)分組。

# 使用summarize()進(jìn)行分組摘要
by_year <- group_by(flights,year,month)
summarise(by_year,delay=mean(arr_delay-dep_delay,na.rm = T))
####查看
(delay_byDay <- group_by(flights,month) %>%summarise(delay_time=mean(dep_delay,na.rm = T))) %>% ggplot(mapping = aes(x=month,y=delay_time))+geom_point()+geom_smooth(se=F)

9. 利用管道符%>%對(duì)數(shù)據(jù)綜合操作：
   • 綜合就是flights %>% group_by(~) %>% summarize(mean(~~,na.rm=T)) %>% filter(~) %>% ggplot(aes())+geom_~()
   • 缺失值：na.rm=T迂猴，缺失值計(jì)算會(huì)都變成缺失值慕淡，可利用filter(!is.na(dep_delay),!is.na(arr_delay))
   • 常用的摘要函數(shù)：n()/sum()/mean()
   • 中位數(shù)median()，分散程度sd()/IQR()/mad()
   • 計(jì)數(shù) n(),計(jì)算唯一值的數(shù)量n_distinct()去重復(fù)后唯一值的計(jì)數(shù)沸毁，count()可快速的計(jì)算峰髓。
   • 邏輯值的計(jì)數(shù) 和 比例：summarize(n_early=sum(dep_time<50)),sum找出大于x的True的數(shù)量，mean會(huì)計(jì)算比例息尺。

# 使用summarize()進(jìn)行分組摘要
by_year <- group_by(flights,year,month)
summarise(by_year,delay=mean(arr_delay-dep_delay,na.rm = T))
####查看
(delay_byDay <- group_by(flights,month) %>%summarise(delay_time=mean(dep_delay,na.rm = T))) %>% ggplot(mapping = aes(x=month,y=delay_time))+geom_point()+geom_smooth(se=F)

### 使用管道組合多種操作
(delay_by_dest <- group_by(flights,dest)%>%summarise(count=n(),delay_time=mean(dep_time,na.rm = T), dist=mean(distance,na.rm = T))) %>% filter(count>20,dest!="HNL") %>% ggplot(mapping = aes(x=dist,y=delay_time))+geom_point(aes(size=count))+geom_smooth(se=F,color="darkblue")

## 管道符 %>%
(delay <- summarise(by_dest,count=n(),dist=mean(distance,na.rm = T),delay=mean(arr_delay,na.rm = T))) ### count=n()統(tǒng)計(jì)分組携兵，就是dest城市的個(gè)數(shù)
delay <- filter(delay,count>20,dest!="HNL")## 篩掉飛行記錄少的，特殊機(jī)場(chǎng)
ggplot(data = delay,mapping = aes(x=dist,y=delay))+geom_point(aes(size=count),alpha=1/3)+geom_smooth(se=F,color="darkblue")
(delay_by_dest <- group_by(flights,dest)%>%summarise(count=n(),delay_time=mean(dep_time,na.rm = T), dist=mean(distance,na.rm = T))) %>% filter(count>20,dest!="HNL") %>% ggplot(mapping = aes(x=dist,y=delay_time))+geom_point(aes(size=count))+geom_smooth(se=F,color="darkblue")
##查看飛機(jī)型號(hào)與延誤時(shí)間的關(guān)系
flights %>% group_by(tailnum) %>%summarise(count=n(),delay_time=mean(arr_delay,na.rm = T)) %>%arrange(delay_time) %>%ggplot(mapping = aes(x=delay_time))+geom_freqpoly(binwidth = 10)
##查看航班數(shù)量 與 飛機(jī)延誤時(shí)間的關(guān)系：航班數(shù)量少時(shí)搂誉，平均延誤時(shí)間的變動(dòng)特別大
delay_time %>% filter(count>25) %>% ggplot(mapping = aes(x=count,y=delay_time))+geom_point(alpha=1/5)
##其他常用的統(tǒng)計(jì)函數(shù)
flights_not_cancelled %>% group_by(dest) %>% summarise(carrier=n())
flights_not_cancelled %>% group_by(dest) %>% summarise(carriers=n_distinct(carrier))
flights_not_cancelled %>% group_by(tailnum) %>% summarise(sum(distance))
 
###exercises
##### 查看哪個(gè)航空公司延誤時(shí)間最長(zhǎng)
flights_not_cancelled %>% group_by(carrier) %>% summarise(count=n(),arr_delay_time=mean(arr_delay)) %>% arrange(desc(arr_delay_time)) %>% ggplot(mapping = aes(x=carrier,y=arr_delay_time))+geom_point(aes(size=count))

10. 取消分組：ungroup()函數(shù)：