Question

我有一个数据框架，其类别为水果，成熟度和均值。如何创建一个运行ttest的for循环，以确定每个水果成熟度的均值差？换句话说，对于苹果，ttest会得出成熟和未成熟苹果之间均值差异的结果。如下表所示。

Answer 1

当您遍历数据中出现的唯一“水果”时，类似这样的方法可能会返回比较“成熟度”的t检验的p值。

## create a vector of the unique fruit in the data; vector of fruit to be tested
fruit<-unique(data$Fruits)
## iterate through your list of unique fruit, testing as you go
for(i in 1:length(fruit)){
  ## subset your data to include only the current fruit to be tested
  df<-filter(data, Fruits==fruit[i])
  ## let the user know which fruit is being tested
  message(fruit[i])
  ## create a vector of the unique ripeness states of the current fruit to be tested
  ripe<-unique(df$Ripeness)
  ## make sure two means exist; ensure there are both ripe and non-ripe values
  if(length(ripe) < 2){
    ## if only one ripeness, let user know and skip to next unique fruit
    message("only one ripeness")
    next
  }
  ## try testing the fruit and return p-value if success
  tryCatch(
    {
      message(t.test(Mean ~ Ripeness, data = df)$p.value)
    },
    ## if error in t-testing return message that there are "not enough observations"
    error=function(cond) {
      message("not enough observations")
    }
  )    
}

我希望这会有所帮助！

Answer 2

假设fruits被编码为类别变量（即应为factor），则可以使用sapply来迭代每个水果的子集数据。在t.test中，我们使用alternative="two.sided"，只是为了强调它的默认设置。

但是，您的数据非常小，Bananas才刚刚成熟。因此，我将使用更大的样本数据集进行演示。

res <- sapply(levels(dat$fruits), function(x) 
  t.test(mean ~ ripeness, dat[dat$fruits %in% x, ], alternative="two.sided")
)
res
#             Apple                     Banana                    Orange                   
# statistic   0.948231                  0.3432062                 0.4421971                
# parameter   23.38387                  30.86684                  16.47366                 
# p.value     0.3527092                 0.7337699                 0.664097                 
# conf.int    Numeric,2                 Numeric,2                 Numeric,2                
# estimate    Numeric,2                 Numeric,2                 Numeric,2                
# null.value  0                         0                         0                        
# stderr      0.8893453                 1.16548                   1.043739                 
# alternative "two.sided"               "two.sided"               "two.sided"              
# method      "Welch Two Sample t-test" "Welch Two Sample t-test" "Welch Two Sample t-test"
# data.name   "mean by ripeness"        "mean by ripeness"        "mean by ripeness"

数据：

set.seed(42)
n <- 1e2
dat <- data.frame(fruits=factor(sample(1:3, n, replace=T),
                                labels=c("Apple", "Banana", "Orange")),
                  ripeness=factor(rbinom(n, 1, .4), labels=c("yes", "no")),
                  mean=round(runif(n)*10))

^{请注意，将来您应该包括一个最小的独立示例，其中应包含适当格式的数据（从不提供图像，请阅读here，以了解如何实现此目的），以及所有您到目前为止已尝试的步骤，因为Stack Overflow不提供编码服务。干杯！}

当R中有两组时，如何使用两个样本的t检验？

2 个答案: