R--组比较：表和可视化

小技巧：R的多行注释：现在Rstudio中选定要注释的区域，然后按ctrl+shif+c可进行多行注释

我们根据我们自己所做的数据和数据框来探索数据：

第一部分构造数据结构：

######数据结构########segVars<-c("age","gender","income","kids","ownHome","subscribe")segVargType<-c("norm","binom","norm","pois",'binom','binom')segNames<-c("Suburb","Urban hip","Travelers","Moving up")segSize<-c(100,50,80,70)#########通过矩阵赋值########segMeans<-matrix(c(40,0.5,55000,2,0.5,0.1,                   24,0.7,21000,1,0.2,0.2,                   58,0.5,64000,0,0.7,0.05,                   36,0.3,52000,2,0.3,0.2),                   ncol=length(segVars),                   byrow = TRUE                 )segSDs<-matrix(c(  5,NA,12000,NA,NA,NA,  2,NA,5000,NA,NA,NA,  8,NA,21000,NA,NA,NA,  4,NA,1000,NA,NA,NA),  ncol = length(segVars),byrow = TRUE)#i.seq<-rep(sqrt(seq(from=2.1,to=6.2,by=1.7)),3)#help("seq")#for( i  in  i.seq){#  print(i)#}#for( i in c("hello"," welcome to"," R world!"))#{cat(i)}#x<-1:5#ifelse(x>1,"hi","bye")######最终客户数据分组数组  seq_along()函数;该函数能给出一个和给定向量some variable  一样长的向量   ##############seg.df<-NULLset.seed(02554)#下面的一句是四组的样泵量取值 取值按照所规定的概率分布去取值for(i in seq_along(segNames)){    cat(i,segNames[i],"\n")  #这里定义了一个this,seg，这是一个临时的矩阵，因为每一次的循环后，这个矩阵边归为NULL，在这段程序的最后的一个操作里面便将这#这个矩阵的临时值通过一个函数不断的放到seg.df之中  this.seg<-data.frame(matrix(NA,nrow = segSize[i],ncol = length(segVars)))##这里的nrow = segSize[i]，一次性产生一个组中的所有数据，比如说第一个数值100，一次性产生100行，6列的矩阵数据值        ##下面就是判定根据不同的匹配值给每列产生数据，一列的数据是100.      for(j in seq_along(segVars)){    if(segVargType[j]=="norm"){            this.seg[,j]<-rnorm(segSize[i],mean = segMeans[i,j],sd=segSDs[i,j])                }else if(segVargType[j]=="pois"){      this.seg[,j]<-rpois(segSize[i],lambda = segMeans[i,j])    }else if(segVargType[j]=="binom"){      this.seg[,j]<-rbinom(segSize[i],size=1,prob = segMeans[i,j])    }else{      stop("Bad segment data type:",segVargType[i])    }      }    ########rbind是取一个向量、矩阵或数据帧参数的序列，并分别由列或行组合起来。这些是带有其他R类方法的泛型函数   seg.df<-rbind(seg.df,this.seg)    }####程序运行完：#1 Suburb #2 Urban hip #3 Travelers #4 Moving up# 为列命名names(seg.df)<-segVarshelp("rep")# rep(x,...)代表复制x中的值# 我们要为列命名，添加类别，然后并把每个二项量变换成因子seg.df$Segment<-factor(rep(segNames,times=segSize))seg.df$ownHome<-factor(seg.df$ownHome,labels = c("ownNo","ownYes"))seg.df$gender<-factor(seg.df$gender,labels = c("Female","male"))seg.df$subscribe<-factor(seg.df$subscribe,labels = c("sunNo","subYes"))summary(seg.df)####### > summary(seg.df)# age           gender        income            kids        ownHome     subscribe        Segment   # Min.   :19.26   Female:157   Min.   : -5183   Min.   :0.00   ownNo :159   sunNo :260   Moving up: 70  # 1st Qu.:33.01   male  :143   1st Qu.: 43731   1st Qu.:0.00   ownYes:141   subYes: 40   Suburb   :100  # Median :39.49                Median : 51997   Median :1.00                             Travelers: 80  # Mean   :41.20                Mean   : 50707   Mean   :1.27                             Urban hip: 50  # 3rd Qu.:47.90                3rd Qu.: 59856   3rd Qu.:2.00                                            # Max.   :80.49                Max.   :114278   Max.   :7.00    # 保存数据到你的电脑，我们以后将对这个数据框进行探索save(seg.df,file = "F:\\segdf-Rintro-Ch5.RData")

# #我们对各组对应的描述统计量# 找到标准行，我们看到："Moving up"类的平均收入，类似的我们对生成的值在进行筛选，加入&  mean(seg.df$income[seg.df$Segment=="Moving up"])mean(seg.df$income[seg.df$Segment=="Moving up"&seg.df$subscribe=="sunNo"])# # 对于上述的过程相当繁琐，更加一般的方式是：用by(data,INDICES,FUN),by()能够按照某个索引变量INDICES# 将数据（data）分成不同的组，然后将函数（FUN）应用到每一组观测值上by(seg.df$income,seg.df$subscribe,mean)by(seg.df$income,seg.df$Segment,mean)# > by(seg.df$income,seg.df$subscribe,mean)# seg.df$subscribe: sunNo# [1] 51389.52# ------------------------------------------------------ # seg.df$subscribe: subYes# [1] 46273.9# > by(seg.df$income,seg.df$Segment,mean)# seg.df$Segment: Moving up# [1] 52109.1# ------------------------------------------------------ # seg.df$Segment: Suburb# [1] 55033.82# ------------------------------------------------------ # seg.df$Segment: Travelers# [1] 62213.94# ------------------------------------------------------ # seg.df$Segment: Urban hip# [1] 21681.93# # 那你也可以用多个因子变量通过list()一起传递给by来进行函数划分by(seg.df$income,list(seg.df$Segment,seg.df$subscribe),mean)# # 对于上面的操作，我们更喜欢用aggregate()来完成，它返回的是一个列表格式aggregate(seg.df$income,list(seg.df$Segment,seg.df$subscribe),mean)# # aggregate(seg.df$income,list(seg.df$Segment,seg.df$subscribe),mean)#     Group.1 Group.2        x# 1 Moving up   sunNo 52163.37# 2    Suburb   sunNo 54942.69# 3 Travelers   sunNo 62746.11# 4 Urban hip   sunNo 22082.11# 5 Moving up  subYes 51891.99# 6    Suburb  subYes 56461.41# 7 Travelers  subYes 58488.77# 8 Urban hip  subYes 20081.19# # 下面的操作，我们看到seg.income.mean就是将第二列的值与seg.df$Segment相匹配，# 然后在seg.df添加一列segIncomeseg.income.mean<-aggregate(seg.df$income,list(seg.df$Segment),mean)seg.df$segIncome<-seg.income.mean[seg.df$Segment,2]library(car)some(seg.df)# > some(seg.df)#          age gender    income kids ownHome subscribe   Segment segIncome# 71  36.22376 Female  53428.06    1   ownNo     sunNo    Suburb  55033.82# 84  34.28161   male  49346.70    2  ownYes    subYes    Suburb  55033.82# 86  35.87205   male  39442.72    1   ownNo     sunNo    Suburb  55033.82# 91  45.19053   male  68937.20    5  ownYes     sunNo    Suburb  55033.82# 171 64.85144 Female  78231.33    0  ownYes     sunNo Travelers  62213.94# 173 64.70641   male  45517.15    0   ownNo    subYes Travelers  62213.94# 184 68.05148   male 104312.45    0  ownYes     sunNo Travelers  62213.94# 185 50.50749 Female  48946.07    0  ownYes    subYes Travelers  62213.94# 202 64.63338   male 113456.80    0   ownNo     sunNo Travelers  62213.94# 290 42.40763   male  52152.04    1   ownNo     sunNo Moving up  52109.10##下面的代码大家敲一遍，然后用上面的some()函数，看看有什么不同seg.income.mean[seg.df$Segment,]seg.df$segIncome<-NULL# # R通过指定的公式来描述变量之间的关系提供标准方式，公式通过（~）操作符将响应变量（左边）和解释变量（右边）# # Y~xy,在aggergate（formula，data,FUN）,下面的是意思是：从数据框seg.df中找到income，将其按照# segment划分成不同的组，然后对每组应用mean(均值函数)#双向量的描述统计量y~x1+X2+....aggregate(income~Segment,data=seg.df,mean)aggregate(income~Segment+ownHome,data = seg.df,mean)agg.data<-aggregate(income~Segment+ownHome,data = seg.df,mean)agg.data[2,3]# aggregate(income~Segment,data=seg.df,mean)#     Segment   income# 1 Moving up 52109.10# 2    Suburb 55033.82# 3 Travelers 62213.94# 4 Urban hip 21681.93# > aggregate(income~Segment+ownHome,data = seg.df,mean)#     Segment ownHome   income# 1 Moving up   ownNo 52249.77# 2    Suburb   ownNo 54932.83# 3 Travelers   ownNo 63188.42# 4 Urban hip   ownNo 21337.59# 5 Moving up  ownYes 51821.64# 6    Suburb  ownYes 55143.21# 7 Travelers  ownYes 61889.12# 8 Urban hip  ownYes 23059.27# # 我们或许还想知道不同的Segment和ownhome层级组合对应的观测函数（frequent）,我们可# 通过table(factor1,factor2)计算单向或者多向频数with(seg.df,table(Segment,ownHome))with(seg.df,table(kids,Segment))# with(seg.df,table(Segment,ownHome))#            ownHome# Segment     ownNo ownYes#   Moving up    47     23#   Suburb       52     48#   Travelers    20     60#   Urban hip    40     10# > with(seg.df,table(kids,Segment))#     Segment# kids Moving up Suburb Travelers Urban hip#    0        13     11        80        17#    1        17     36         0        17#    2        18     22         0        11#    3        13     19         0         4#    4         5      7         0         1#    5         3      3         0         0#    6         0      2         0         0#    7         1      0         0         0#下面四个方法的就是对某个类中的所有孩子的个数加起来得到样本的小孩总数。#实现方式一xtabs(kids~Segment,data=seg.df)#实现方式二aggregate(kids~Segment,data = seg.df,sum)#实现方式三seg.tab<-with(seg.df,table(kids,Segment))apply(seg.tab*0:7,2,sum)#实现方式四colSums(seg.tab*0:7)require(lattice)# ##下面的就是跟上面的类似，比如说我们要对每个类中订阅者做比例图，# # 了解不同类的订阅情况 histogram(formula,data,type)# 下面的三行代码分别对应图1,2,3histogram(~subscribe|Segment,data = seg.df)histogram(~subscribe|Segment,data = seg.df,          type = "count",layout=c(4,1),col=c("burlywood","darkolivegreen"))histogram(~subscribe|Segment+ownHome,data = seg.df)# ###prop.table（table,margin）命令，如果你将prop.table（，margin=...）作用于table()# 上可以得到每个单元格数值对整个表格之和的比例，或对行求和的比例，或者对列的求和比例（margin=2）prop.table(table(seg.df$subscribe,seg.df$Segment),margin = 2)# > prop.table(table(seg.df$subscribe,seg.df$Segment),margin = 2)#         #          Moving up Suburb Travelers Urban hip#   sunNo      0.800  0.940     0.875     0.800#   subYes     0.200  0.060     0.125     0.200# # # 大家对多获得数据画图，就是对有房的人进行订阅统计，大家见图4barchart(prop.table(table(seg.df$subscribe,seg.df$Segment),margin = 2)[2,],         xlab = "subscriber proportion by segment",col="darkolivegreen")#下面的例子是对前几个上面例子综合使用seg.means<-aggregate(income~Segment,data =seg.df,mean)library(lattice)barchart(income~Segment,data = seg.means,col="grey")#图6显示seg.income.agg<-aggregate(income~Segment+ownHome,data = seg.df,mean)barchart(income~Segment,data = seg.income.agg,groups=ownHome,auto.key=TRUE,         par.setting=simpleTheme(col=terrain.colors(2)))#图7 箱线图boxplot(income~Segment,data = seg.df,yaxt="n",ylab="Incom($k)")ax.seq<-seq(from=0,to=120000,by=20000)axis(side=2,at=ax.seq,labels = paste(ax.seq/1000,"k",seq=""),las=1)bwplot(Segment~income,data = seg.df,horizontal = TRUE,xlab="Income")#lattic中有个更好的绘制箱线图的bwplot,bwplot()使用与常规方向相反的模式公式表达。bwplot(Segment~income|ownHome,data = seg.df,horizontal = TRUE,xlab="Income")