選擇 output 用於 Tukey 的 HSD 在 R 中的雙向 ANOVA

Question

我有一個包含多個變量的大型數據集。 我需要使用 Tukey HSD 進行雙向方差分析，然后進行事后成對多重比較。

我的前 25 個條目的數據頭是這樣的：

> head(my_data2, 25 )
   CellType variable     value
1     Cell1       W1 18.780294
2     Cell1       W1 13.932397
3     Cell1       W1 20.877093
4     Cell1       W1  9.291295
5     Cell1       W1 10.939570
6     Cell1       W1 12.236713
7     Cell1       W1 13.810722
8     Cell1       W1 23.944473
9     Cell1       W1 17.355429
10    Cell1       W1 18.248215
11    Cell2       W1 17.988200
12    Cell2       W1 15.427909
13    Cell2       W1 21.839687
14    Cell2       W1 22.322325
15    Cell2       W1 12.535762
16    Cell2       W1 12.743278
17    Cell2       W1 15.007214
18    Cell2       W1 12.054787
19    Cell2       W1 15.639977
20    Cell2       W1 16.006960
21    Cell3       W1 17.452199
22    Cell3       W1 23.280391
23    Cell3       W1  7.902728
24    Cell3       W1  8.353992
25    Cell3       W1 24.360250

我做方差分析

#ANOVA
my_data2$CellType <- as.factor(my_data2$CellType)
my_ANOVA = aov(value ~ CellType + variable + CellType:variable, data = my_data2)
summary(my_ANOVA)

然后事后

my_posthoc =TukeyHSD(my_ANOVA, which = "CellType:variable") 
my_posthoc

到目前為止一切都還好，但是我的 posthoc 的 output 包括所有成對比較，這給了我們超過 2200 行的大量看跌。 例如我的輸出是這樣的：

> my_posthoc
  Tukey multiple comparisons of means
    95% family-wise confidence level

Fit: aov(formula = value ~ CellType + variable + CellType:variable, data = my_data2)

$`CellType:variable`
                       diff          lwr         upr     p adj
Cell2:W1-Cell1:W1   0.21499 -29.46177884  29.8917588 1.0000000
Cell3:W1-Cell1:W1   0.88234 -28.79442884  30.5591088 1.0000000
Cell4:W1-Cell1:W1   1.24301 -28.43375884  30.9197788 1.0000000
Cell5:W1-Cell1:W1   1.61684 -28.05992884  31.2936088 1.0000000
Cell6:W1-Cell1:W1   0.65009 -29.02667884  30.3268588 1.0000000
Cell7:W1-Cell1:W1   1.08223 -28.59453884  30.7589988 1.0000000
Cell1:W2-Cell1:W1   9.00094 -20.67582884  38.6777088 1.0000000
Cell2:W2-Cell1:W1  27.62765  -2.04911884  57.3044188 0.1249342
Cell3:W2-Cell1:W1  29.40077  -0.27599884  59.0775388 0.0570151
Cell4:W2-Cell1:W1  28.84731  -0.82945884  58.5240788 0.0736530
Cell5:W2-Cell1:W1  42.51407  12.83730116  72.1908388 0.0000144
Cell6:W2-Cell1:W1  30.78610   1.10933116  60.4628688 0.0288235
Cell7:W2-Cell1:W1  27.62966  -2.04710884  57.3064288 0.1248307
Cell1:W3-Cell1:W1  20.95847  -8.71829884  50.6352388 0.7816085
Cell2:W3-Cell1:W1  42.50116  12.82439116  72.1779288 0.0000146
Cell3:W3-Cell1:W1  47.07037  17.39360116  76.7471388 0.0000004
Cell4:W3-Cell1:W1  47.26760  17.59083116  76.9443688 0.0000003
Cell5:W3-Cell1:W1  64.08026  34.40349116  93.7570288 0.0000000
Cell6:W3-Cell1:W1  53.90284  24.22607116  83.5796088 0.0000000

最后說：

[ reached getOption("max.print") -- omitted 2290 rows ]

但是我只對每個變量內的比較感興趣，而不是它們之間的比較。 以上述 output 為例，我只需要Cell1:W1-Cell2:W1 。 都在同一個變量w1中。 或例如Cell6:W3-Cell1:W3 。 我對Cell6:W3-Cell6:W1不感興趣

我該如何指定？ 謝謝

Answer 1

我采取了簡單的誠實方式，將術語（行名）分成四個部分並過濾它們。

library(dplyr); library(tibble); library(purrr)  # OR library(tidyverse)  # EDITED

my_posthoc2 <- my_posthoc %>% 
  pluck("CellType:variablen") %>%             # get element of list
  as_tibble(rownames = "Term") %>%            # convert to tibble
  separate(Term,                              # separate terms by - and :
           into = c("LL", "LR", "RL", "RR"), 
           sep = "-|:", 
           remove = FALSE) 

my_posthoc2 %>% 
  filter(LR == "W1", RR == "W1")  # get Cell1:W1-Cell2:W1

Answer 2

由於您指定“我只對每個變量內的比較感興趣，而不是它們之間的比較”，因此您不需要包含交互項CellType:variable

您可以將 model 重寫為：

my_ANOVA = aov(value ~ CellType + variable, data = my_data2)