CODEWITHSUNDEEP
rstatisticspermutation
Remi.b
Remi.b

Reputation: 18219

Summarizing list of pairwise statistical tests

Intro

After performing an ANOVA, we often perform a series of pairwise comparisons. When the number of groups is relatively large, the number of pairwise comparison becomes very big (it is its triangular number). On graphs, it is common to indicate groups that are not significantly different by marking them with the same letter and it would therefore be handy to have a function that would produce these letters.

Goal

I am trying to build a function in R that would take in input:

.. and would output:

Example

set.seed(10)
d = data.frame(
    y=c(rnorm(100,10),rnorm(100,10),rnorm(100,8.5),rnorm(100,8)),
    x=rep(c("High","MidHigh", "MidLow","Low"), each=100)
)

z=TukeyHSD(aov(y~x,data=d))
z
  Tukey multiple comparisons of means
    95% family-wise confidence level

Fit: aov(formula = y ~ x, data = d)

$x
                      diff         lwr        upr     p adj
Low-High       -1.66461230 -2.02288676 -1.3063378 0.0000000
MidHigh-High    0.04158636 -0.31668810  0.3998608 0.9906564
MidLow-High    -1.33469921 -1.69297367 -0.9764247 0.0000000
MidHigh-Low     1.70619866  1.34792420  2.0644731 0.0000000
MidLow-Low      0.32991309 -0.02836138  0.6881876 0.0834543
MidLow-MidHigh -1.37628557 -1.73456004 -1.0180111 0.0000000

And here is what the function fun would do

fun(Tukey=z, alpha=0.05)
High MidHigh  MidLow     Low 
"A"     "A"     "B"     "B" 

fun(Tukey=z, alpha=0.1)
High MidHigh  MidLow     Low 
"A"     "A"     "B"     "C"

Upvotes: 0

Views: 171

Answers (1)

Aaron - mostly inactive
Aaron - mostly inactive

Reputation: 37754

Try the multcompView package.

library(multcompView)
multcompLetters(z$x[,4])
##    Low MidHigh  MidLow    High 
##    "a"     "b"     "a"     "b"

Upvotes: 3

Related Questions