> tdata <- data.frame(names=rownames(tdata),tdata)行名作為第一列
> gather(tdata,key="Key",value="Value",cyl:disp,mpg)創(chuàng)key列和value列，cyl和disp放在一列中
-號減去不需要轉換的列
> spread(gdata,key="Key",value="Value")
根據value將key打散開 與unite函數對立
separate(df,col=x,into=c("A","B"))將數據框的列分割
unite(x,col="AB",A,B,sep='.')

> dplyr::filter(iris,Sepal.Length>7)條件過濾
> dplyr::distinct(rbind(iris[1:10,],iris[1:15,]))去除重復行
> dplyr::slice(iris,10:15)切片
> dplyr::sample_n(iris,10)隨機10行
> dplyr::sample_frac(iris,0.1)按比例隨機選取
> dplyr::arrange(iris,Sepal.Length)排序
dplyr::arrange(iris,desc(Sepal.Length))降序
> select(starwars,height)選取
> summarise(iris,avg=mean(Sepal.Length))

%>%鏈式操作符，管道 ctrl+shift+m
> iris %>% group_by(Species)
> dplyr::group_by(iris,Species)
> iris %>% group_by(Species) %>% summarise(avg=mean(Sepal.Width)) %>% arrange(avg)
> dplyr::mutate(iris,new= Sepal.Length+Petal.Length)相加總和
> dplyr::left_join(a,b,by="x1")
> dplyr::right_join(a,b,by="x1")
> dplyr::full_join(a,b,by="x1")
> dplyr::semi_join(a,b,by="x1")交集部分
> dplyr::anti_join(a,b,by="x1")補集部分
> intersect(first,second)交集
> dplyr::union_all(first,second)并集
> dplyr::union(first,second)非冗余并集
> setdiff(first,second)補集

heatmap輸入矩陣
lm輸入數據框
plot向量和向量-散點圖，向量和因子-條形圖
cbind,rbind矩陣或數據框
sum,mean,sd,range,median,sort,order向量
main 字符串不能為向量
na.rm true和false
axis side參數1到4
fig 包含四個元素向量
> plot(c(1:20),c(seq(1,89,length.out=20)),type="l",lty=1)實線
> plot(c(1:20),c(seq(1,89,length.out=20)),type="l",lty=2)虛線

> x <- rnorm(n=100,mean=15,sd=2)生成100個平均數為15方差為2的隨機數
> qqnorm(x)
set.seed(666) runif(50)綁定隨機數
dgama(c(1:9),shape=2,rate=1)生成密度gama分布；隨機數

summary()
fivenum()
Hmisc describe()
pastecs stat.desc() basic=T norm=T
psych describe() trim=0.1去除最低最高10%
> aggregate(Cars93[c("Min.Price","Price","Max.Price"," MPG.city")],by=list(Manufacturer=Cars93$Manufacturer),mean)字符串型 返回一個統計函數
doBy > summaryBy(mpg+hp+wt~am,data=myvars,FUN = mean)
psych describe.by(myvars,list(am=mtcars$am))分組統計
describeBy(myvars,list(am=mtcars$am))詳細信息

> table(cut(mtcars$mpg,c(seq(10,50,10))))頻數統計
> prop.table(table(mtcars$cyl))頻數占比
> table(Arthritis$Treatment,Arthritis$Improved)
> with(data=Arthritis,(table(Treatment,Improved)))省略數據集的名字
> xtabs(~Treatment+Improved,data=Arthritis)根據類別統計頻數
> margin.table(x,1/2)總和
> addmargins(x)將總和添加到原表中
> ftable(y)評估式類元表

原假設：不變 備擇假設：變化
P值越小越能實現
> mytable <- table(Arthritis$Treatment,Arthritis$Improved)
> chisq.test(mytable)卡方獨立性檢驗
> fisher.test(mytable)精確獨立檢驗
> mantelhaen.test(mytable)
> mytable <- xtabs(~Treatment+Sex+Improved,data=Arthritis)
> mantelhaen.test(mytable)

> cor(state.x77) > cor(x,y)
> cov(state.x77)
偏相關
ggm
> pcor(c(1,5,2,3,6),cov(state.x77))
> cor.test(state.x77[,3],state.x77[,5])
psych
> corr.test(state.x77)
> x <- pcor(c(1,5,2,3,6),cov(state.x77))
> pcor.test(x,3,50)
MASS
> t.test(Prob~So,data=UScrime)

散點圖 x、y
直方圖 因子
熱力圖 數據矩陣
象限圖 因子、向量
> plot(women$height~women$weight)關聯圖
> fit <- lm(height~weight,data=women)
> plot(fit)
S3 par/plot/summary
> plot(as.factor(mtcars$cyl),col=c("red","yellow","blue"))

> mystats <- function(x,na.omit=FALSE){
+ if(na.omit)
+ x <- x[!is.na(x)]
+ m <- mean(x)
+ n <- length(x)
+ s <- sd(x)
+ skew <- sum((x-m^3/s^3))/n
+ kurt <- sum((x-m^4/s^4))/n-3
+ return(c(n=m,mean=m,stdev=s,skew=skew,kurtosis=kurt))
+ }

> i=1;while (i<=10){print("Hello,World");i=i+2;}
for(i in 1:10){print("Hello,World")}
> ifelse(score>60,print("PASS"),print("FAIL")

> fit <- lm(weight~height,data=women)
> summary(fit)
> coefficients(fit)
> confint(fit,level=)置信區(qū)間，默認95%
> fitted(fit)擬合模型預測值
源數據-預測值=殘差residuals()
> predict(fit,women1)根據結果對新數據進行預測
殘差擬合圖，正態(tài)分布圖，大小位列圖，殘差影響圖
plot(women$height,women$weight)
abline擬合曲線
> fit2 <- lm(weight~height+I(height^2),data=women)增加二次項
> lines(women$height,fitted(fit2),col="red")
將點連成線，根據擬合曲線
Pr(>|t|)估計系數為0假設的概率，小于0.05
Residual standard error殘差越小越好
Multiple R-squared擬合值越大越好，解釋數據量
F-statistic模型是否顯著，越小越好

AIC比較回歸值擬合度結果
MASS
stepAIC逐步回歸法
leaps
regsubsets全子集回歸法

> par(mfrow=c(2,2)) plot四幅圖顯示在同個畫面
抽樣驗證法
500個數據進行回歸分析，predict對剩下500個預測，比較殘差值

> library(multcomp)
> attach(cholesterol)
> table(trt)
> aggregate(response,by=list(trt),FUN=mean) 分組統計平均值查看效果最好因子
> fit <- aov(response~ trt,data=cholesterol) 方差分析
> summary(fit) 看統計結果，方差結果看F值 越大組間差異越顯著、P值衡量F值越小越可靠

> attach(litter)
> aggregate(weight,by=list(dose),FUN=mean)
> fit <- aov(weight~gesttime+dose,data=litter)
> summary(fit)

> attach(ToothGrowth)
> xtabs(~supp+dose)統計頻率
> aggregate(len,by=list(supp,dose),FUN=mean)劑量越小兩者差別越明顯
> ToothGrowth$dose <- factor(ToothGrowth$dose)
> fit <- aov(len ~ supp*dose,data=ToothGrowth)
> summary(fit)

> interaction.plot(dose,supp,len,type="b",
col=c("red","blue"),pch=c(16,18),
main = "Interaction between Dose and Supplement Type")

> library(MASS)
> attach(UScereal)
> shelf <- factor(shelf)
> aggregate(cbind(calories,fat,sugars),by=list(shelf),FUN=mean)
> summary.aov(fit)每組測量值不同，差異結果顯著

> pwr.f2.test(u=3,sig.level=0.05,power=0.9,f2=0.0769)假設顯著性水平為0.05，在90%置信水平下至少需要184個樣本
pwr.anova.test(k=2,f=0.25,sig.level=0.05,power=0.9) 2組效率為0.25顯著性水平為0.05，功效水平為90，結果為86*2

> data(breslow.dat,package = "robust")
> summary(breslow.dat)
> attach(breslow.dat)
fit <- glm(sumY~Base + Trt +Age，data=breslow.dat,family=poisson(link="log")) 廣義線性模型擬合泊松回歸 響應變量

> data(Affairs,package="AER")
> summary(Affairs)
> table(Affairs$affairs)
> prop.table(table(Affairs$affairs)）
> prop.table(table(Affairs$gender))
> Affairs$ynaffair[Affairs$affairs>0] <- 1
> Affairs$ynaffair[Affairs$affairs==0] <- 0
> Affairs$ynaffair <- factor(Affairs$ynaffair,levels=c(0,1),labels=c("No","Yes"))
> table(Affairs$ynaffair)
> attach(Affairs )
> fit <- glm(ynaffair~gender+age+yearsmarried+children+religiousness+education+occupation+rating,data=Affairs,family=binomial())
> summary(fit)
> fit1 <- glm(ynaffair~age+yearsmarried+religiousness+rating,data=Affairs,family=binomial())
> summary(fit1)
> anova(fit,fit1,test="Chisq")

> library(psych)
> fa.parallel(USJudgeRatings,fa="pc",n.iter=100)直線與X符號生成值大于一和100次模擬的平行分析
CPU
> pc <- principal(USJudgeRatings,nfactors=1,rotate="none",scores=FALSE)/scores=T pc1包含成分整合，觀測變量與主成分的相關系數，h2指成分公因子的方差，主成分對每個變量的方差解釋度，u2指方差無法被主成分解釋的比例，SSloadings特定主成分相關聯的標準化后的方差值，proportion var每個主成分對相關值的解釋程度

> library(psych)
> options(digits=2)
> covariances <- ability.cov$cov
> correlations <- cov2cor(covariances)
> fa.parallel(correlations,fa="both",n.obs=112,n.iter=100)
> fa.varimax <- fa(correlations,nfactors=2,rotate="varimax",fm="pa")
> fa.promax <- fa(correlations,nfactors=2,rotate="promax",fm="pa")
factor.plot(fa.promax,labels=rownames(fa.promax$loadings))
fa.diagram(fa.varimax,simple=FALSE)
fa<-fa(correlations,nfactors=2,rotate="none",fm="pa",score=TRUE)
fa$weight

library(arules)
data(Groceries)
> fit <- apriori(Groceries,parameter=list(support=0.01,confidence=0.5))
> inspect(fit)