Wilcoxon循环测试

debugcn 发表于 Dev

我想对我的数据（AllData）全部执行整数的Wilcoxon测试。这是我的原始数据：

Date v1 v2 v3 v4 v5 v6 v7 v8
1   2014-01-05  39  4   84  75  41  6   83  610
2   2014-01-12  40  6   86  77  44  6   84  765
3   2014-01-19  39  5   82  73  40  6   81  713
4   2014-01-26  37  5   100 71  39  6   90  685
5   2014-02-02  39  5   83  70  37  5   79  601
6   2014-02-09  44  6   82  78  40  6   78  535
7   2014-02-16  41  5   76  76  40  7   78  582
8   2014-02-23  40  5   74  72  42  6   81  568
9   2014-03-02  35  4   81  71  39  6   78  502

到目前为止，这是我的编码，可以正常工作

#calculate basefitMAE 
basefit.model1 <- Arima(AllData$v8,order=c(0,1,2)) 
inSampleBaseFitMAE <- mean(abs(basefit.model1$residuals))

#calculate advancedfitMAE
AllModel <- AllData[c(2:8)]
allfit <- data.frame(inSampleAdvancedFitMAE = rep(NA, length(AllModel)))
for (i in seq_along(AllModel[1,])) {
  advancedfit.model <- Arima(AllData$x1,order=c(0,1,2),xreg=AllModel[,i])
  allfit$inSampleAdvancedFitMAE[i] <- mean(abs(advancedfit.model$residuals))
}

allfit <- cbind(allfit,inSampleBaseFitMAE)

#measure relative MAE 
allfit$relativeMAE <- (allfit$inSampleAdvancedFitMAE)/(allfit$inSampleBaseFitMAE)

Wilcoxon的基本公式是：

wilcox.test(abs(basefit.model$residuals),abs(advancedfit.model$residuals),paired=TRUE)

现在，我只需要在循环中进行Wilcoxon测试。但是我对如何执行此操作感到困惑，因为我没有所有数据的绝对残差，而目前只有平均残差。

笨狼

我需要创建更多数据来进行拟合，因此：

AllData <- data.frame(Date=as.Date("2014-01-05") + seq(0,700,by=7),
matrix(rnbinom(101*8,mu=10,size=1),ncol=8))

colnames(AllData)[2:9] <- paste0("v",1:8)

希望我对您有所帮助，并注意Arima(AllData$x1..)您发布的代码中的错字，显示的表中不存在x1。

basefit.model1 <- Arima(AllData$v8,order=c(0,1,2)) 
inSampleBaseFitMAE <- mean(abs(basefit.model1$residuals))

allfit <- data.frame(var = colnames(AllData)[2:8],
inSampleAdvancedFitMAE =NA,p.value=NA,
inSampleBaseFitMAE=inSampleBaseFitMAE,stringsAsFactors=FALSE)

for (i in seq_along(allfit$var)) {
  advancedfit.model <- Arima(AllData$v8,order=c(0,1,2),xreg=AllData[,allfit$var[i]])
  allfit$inSampleAdvancedFitMAE[i] <- mean(abs(advancedfit.model$residuals))
  test = wilcox.test(advancedfit.model$residuals,basefit.model1$residuals,paired=TRUE)
  allfit$p.value[i] <- test$p.value
}

allfit$relativeMAE <- (allfit$inSampleAdvancedFitMAE)/(allfit$inSampleBaseFitMAE)
allfit

  var inSampleAdvancedFitMAE     p.value inSampleBaseFitMAE relativeMAE
1  v1               7.919351 0.025809584           7.955079   0.9955089
2  v2               7.859983 0.954075356           7.955079   0.9880459
3  v3               7.968860 0.165886368           7.955079   1.0017324
4  v4               7.796261 0.128247572           7.955079   0.9800356
5  v5               7.940206 0.002050978           7.955079   0.9981304
6  v6               7.960803 0.122403761           7.955079   1.0007195
7  v7               7.929170 0.111342188           7.955079   0.9967432

本文收集自互联网，转载请注明来源。

如有侵权，请联系[email protected] 删除。