具有测试数据的统计测试

Question

如果我在插入符号中使用两种方法（NN和KNN），然后我想提供显着性检验，那我该怎么做wilcoxon检验。

我提供了如下数据样本

structure(list(Input = c(25, 193, 70, 40), Output = c(150, 98, 
        27, 60), Inquiry = c(75, 70, 0, 20), File = c(60, 36, 12, 12), 
        FPAdj = c(1, 1, 0.8, 1.15), RawFPcounts = c(1750, 1902, 535, 
        660), AdjFP = c(1750, 1902, 428, 759), Effort = c(102.4, 
        105.2, 11.1, 21.1)), row.names = c(NA, 4L), class = "data.frame")

    d=readARFF("albrecht.arff") 
    index <- createDataPartition(d$Effort, p = .70,list = FALSE)
    tr <- d[index, ]
    ts <- d[-index, ] 

    boot <- trainControl(method = "repeatedcv", number=100)

         cart1 <- train(log10(Effort) ~ ., data = tr,
                        method = "knn",
                        metric = "MAE",
                        preProc = c("center", "scale", "nzv"),
                        trControl = boot)

           postResample(predict(cart1, ts), log10(ts$Effort))

           cart2 <- train(log10(Effort) ~ ., data = tr,
                          method = "knn",
                          metric = "MAE",
                          preProc = c("center", "scale", "nzv"),
                          trControl = boot)

           postResample(predict(cart2, ts), log10(ts$Effort))

如何在此处执行wilcox.test()。

    Warm regards

Answer 1

这对您有用吗？

library(caret)
df <- structure(list(Input = c(25, 193, 70, 40), Output = c(150, 98, 
                                                      27, 60), Inquiry = c(75, 70, 0, 20), File = c(60, 36, 12, 12), 
               FPAdj = c(1, 1, 0.8, 1.15), RawFPcounts = c(1750, 1902, 535, 
                                                           660), AdjFP = c(1750, 1902, 428, 759), Effort = c(102.4, 
                                                                                                             105.2, 11.1, 21.1)), row.names = c(NA, 4L), class = "data.frame")

# not enough data points in df for ML: increase the number of df rows X10
d <- df[rep(seq_len(nrow(df)), 10), ]

index <- createDataPartition(d$Effort, p = .70,list = FALSE)
tr <- d[index, ]
ts <- d[-index, ] 

boot <- trainControl(method = "repeatedcv", number=100)

cart1 <- train(log10(Effort) ~ ., data = tr,
               method = "knn",
               metric = "MAE",
               preProc = c("center", "scale", "nzv"),
               trControl = boot)

# save the output to "model_predictions_1"
model_predictions_1 <- postResample(predict(cart1, ts), log10(ts$Effort))

cart2 <- train(log10(Effort) ~ ., data = tr,
               method = "knn",
               metric = "MAE",
               preProc = c("center", "scale", "nzv"),
               trControl = boot)

# save the output to "model_predictions_2"
model_predictions_2 <- postResample(predict(cart2, ts), log10(ts$Effort))

# test model_predictions_1 vs model_predictions_2
wilcox.test(model_predictions_1, model_predictions_2, exact = FALSE)