'Calculate partial eta-squared with type 3 sum of square in r
I have ran a 2 X 2 X 2 mixed ANOVA using ezANOVA and type 3 sum of squared in r.
The code looks like
ezANOVA(data = D, between = condition, within = c(Notation,Operation), dv = Acc, wid = ID,type=3)
The output does not include the sum of square and the effect size was the generalized eta-squared. I am not sure how to calculate the partial eta-squared with type 3 sum of square in r.
I have tried to use the aov() function and eta_squared（） function from package effectsize, but the aov() function uses type 1 sum of square and so the effect size is different from the type 3 sum of square effect size.
Thus, I am wondering if there is any way to calculate the partial eta squared for a 3-way mixed ANOVA using type 3 sum of square in R.
Thank you in advance for your help
Solution 1:[1]
Short of manually calculating partial eta squared, I wasn't able to find a function that worked with ezANOVA. I want to point out that the column labeled ges is the generalized eta squared (not partial, though).
However, I do have a method that will work for using SS 3 from the package jmv, along with both within and between ANOVA, while providing partial eta squared. It's a bit more of a mouthful to put together the function, as well. I added tons of options that you don't have outlined in your function in the question. I did this because this package's help isn't all that helpful. You definitely don't need to use all of these parameters, but at least you'll know what the package is expecting if you do use these options.
Your question isn't reproducible. I started by creating some arbitrary data to work with.
library(jmv)
# some fake data to work with
set.seed(253)
df1 <- data.frame(x = rnorm(200, 50, 3),
y = rnorm(200, 25, 5),
z = rnorm(200, 1.5, .1),
direc = as.factor(rep(c("left","right"), times = 100)))
Next the repeated measures + between ANOVA:
fit = anovaRM(data = df1,
ss = "3", # type of SS (1, 2, or 3)
bs = list("direc"), # between subjects
bsTerms = list("direc"), # between subjects
rm = list(list(label = "tests", # within subjects
levels = c("pretest","mid","posttest"))),
# can use levels(data$factor) if easier
# does not have to be a real variable**
rmCells = list(list(measure = "x", # continuous value
cell = "pretest"), # group label
list(measure = "y", # continuous value
cell = "mid"), # group label
list(measure = "z", # continuous value
cell = "posttest")), # group
rmTerms = list("tests"), # grouping variable/within measures
emMeans = list(list("tests","direc")), # all grouping vars (em tables)
emmPlots = T, # show emm plot
emmTables = T, # show emm tables
effectSize = "partEta", # use partial eta (multi options, see help)
spherTests = T, # use Mauchley test
spherCorr = "GG", # Greenhouse (multi options`, see help)
leveneTest = T, # check homogeneity (p > .05 = good**)
qq = T, # plot normality validation qq plot
postHocCorr = "tukey") # use TukeyHSD
This is the type of output you'll see when you call fit (or whatever you name your ANOVA object).
#
# REPEATED MEASURES ANOVA
#
# Within Subjects Effects
# ??????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????
# Sphericity Correction Sum of Squares df Mean Square F p ?²-p
# ??????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????
# tests Greenhouse-Geisser 235493.0127 1.399431 168277.62891 9808.561967 < .0000001 0.9802130
# tests:direc Greenhouse-Geisser 105.1735 1.399431 75.15443 4.380599 0.0247609 0.0216454
# Residual Greenhouse-Geisser 4753.7668 277.087435 17.15620
# ??????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????
# Note. Type 3 Sums of Squares
#
#
# Between Subjects Effects
# ??????????????????????????????????????????????????????????????????????????????????????????
# Sum of Squares df Mean Square F p ?²-p
# ??????????????????????????????????????????????????????????????????????????????????????????
# direc 22.01519 1 22.01519 1.954001 0.1637204 0.0097723
# Residual 2230.81167 198 11.26673
# ??????????????????????????????????????????????????????????????????????????????????????????
# Note. Type 3 Sums of Squares
#
#
# ASSUMPTIONS
#
# Tests of Sphericity
# ???????????????????????????????????????????????????????????????????????????????
# Mauchly's W p Greenhouse-Geisser ? Huynh-Feldt ?
# ???????????????????????????????????????????????????????????????????????????????
# tests 0.5708482 < .0000001 0.6997157 0.7031690
# ???????????????????????????????????????????????????????????????????????????????
#
#
# Homogeneity of Variances Test (Levene's)
# ???????????????????????????????????????????????
# F df1 df2 p
# ???????????????????????????????????????????????
# x 1.653217e-4 1 198 0.9897542
# y 0.42682247 1 198 0.5143102
# z 0.01824029 1 198 0.8927043
# ???????????????????????????????????????????????
#
#
# ESTIMATED MARGINAL MEANS
#
# TESTS:DIREC
#
# Estimated Marginal Means - tests:direc
# ???????????????????????????????????????????????????????????????????????????
# direc tests Mean SE Lower Upper
# ???????????????????????????????????????????????????????????????????????????
# left pretest 50.224630 0.307314811 49.618600 50.830660
# mid 24.048471 0.508157857 23.046375 25.050567
# posttest 1.499185 0.009470430 1.480509 1.517860
# right pretest 49.818121 0.307314811 49.212091 50.424151
# mid 25.590657 0.508157857 24.588561 26.592753
# posttest 1.512816 0.009470430 1.494140 1.531492
# ???????????????????????????????????????????????????????????????????????????
#
These are the two plots you'll see (or a variation of these: qq and emm).
This is really an amazing package, but it isn't very self-explanatory. If you have any questions, leave a comment.
Sources
This article follows the attribution requirements of Stack Overflow and is licensed under CC BY-SA 3.0.
Source: Stack Overflow
| Solution | Source |
|---|---|
| Solution 1 | Kat |