Mixed ANOVA - Domain General Vars
2026-02-03
Setup
library(tidyverse)
library(rstatix)
library(emmeans)
library(effectsize)
library(afex)
library(car)
options(scipen = 999)
afex::set_sum_contrasts()Data
# Data file is in parent of knit/ (eohi3/eohi3.csv)
df <- read.csv(
"/home/ladmin/Documents/DND/EOHI/eohi3/eohi3.csv",
stringsAsFactors = FALSE,
check.names = FALSE,
na.strings = "NA"
)
between_vars <- c("perspective", "temporalDO")
within_vars <- c(
"past_pref_DGEN", "past_pers_DGEN", "past_val_DGEN",
"fut_pref_DGEN", "fut_pers_DGEN", "fut_val_DGEN"
)
missing_vars <- setdiff(c(between_vars, within_vars, "pID"), names(df))
if (length(missing_vars) > 0) {
stop(paste("Missing required variables:", paste(missing_vars, collapse = ", ")))
}
anova_data <- df %>%
select(pID, all_of(between_vars), all_of(within_vars)) %>%
filter(
!is.na(perspective), perspective != "",
!is.na(temporalDO), temporalDO != ""
)Long format
long_data <- anova_data %>%
pivot_longer(
cols = all_of(within_vars),
names_to = "variable",
values_to = "DGEN_SCORE"
) %>%
mutate(
time = case_when(
grepl("^past_", variable) ~ "past",
grepl("^fut_", variable) ~ "fut",
TRUE ~ NA_character_
),
domain = case_when(
grepl("_pref_DGEN$", variable) ~ "pref",
grepl("_pers_DGEN$", variable) ~ "pers",
grepl("_val_DGEN$", variable) ~ "val",
TRUE ~ NA_character_
)
) %>%
mutate(
TIME = factor(time, levels = c("past", "fut")),
DOMAIN = factor(domain, levels = c("pref", "pers", "val")),
perspective = factor(perspective),
temporalDO = factor(temporalDO),
pID = factor(pID)
) %>%
select(pID, perspective, temporalDO, TIME, DOMAIN, DGEN_SCORE) %>%
filter(!is.na(DGEN_SCORE))Descriptive statistics
desc_stats <- long_data %>%
group_by(perspective, temporalDO, TIME, DOMAIN) %>%
summarise(
n = n(),
mean = round(mean(DGEN_SCORE), 5),
variance = round(var(DGEN_SCORE), 5),
sd = round(sd(DGEN_SCORE), 5),
median = round(median(DGEN_SCORE), 5),
q1 = round(quantile(DGEN_SCORE, 0.25), 5),
q3 = round(quantile(DGEN_SCORE, 0.75), 5),
min = round(min(DGEN_SCORE), 5),
max = round(max(DGEN_SCORE), 5),
.groups = "drop"
)
# Show all rows and columns (no truncation)
options(tibble.width = Inf)
print(desc_stats, n = Inf)## # A tibble: 24 × 13
## perspective temporalDO TIME DOMAIN n mean variance sd median q1
## <fct> <fct> <fct> <fct> <int> <dbl> <dbl> <dbl> <dbl> <dbl>
## 1 other future past pref 122 3.74 7.19 2.68 4 1
## 2 other future past pers 122 3.49 7.56 2.75 3 1
## 3 other future past val 122 3.72 7.64 2.76 3 1
## 4 other future fut pref 122 3.69 6.65 2.58 3 2
## 5 other future fut pers 122 3.29 7.03 2.65 3 1
## 6 other future fut val 122 3.43 7.59 2.75 3 1
## 7 other past past pref 139 3.93 7.56 2.75 4 2
## 8 other past past pers 139 3.79 8.12 2.85 3 1
## 9 other past past val 139 3.30 7.26 2.69 3 1
## 10 other past fut pref 139 3.60 7.53 2.74 3 1
## 11 other past fut pers 139 3.50 8.31 2.88 3 1
## 12 other past fut val 139 3.42 9.24 3.04 3 1
## 13 self future past pref 138 3.88 7.39 2.72 3.5 2
## 14 self future past pers 138 3.73 7.48 2.74 3 1
## 15 self future past val 138 3.65 7.75 2.78 3 1
## 16 self future fut pref 138 3.92 6.86 2.62 4 2
## 17 self future fut pers 138 3.72 7.66 2.77 3 1
## 18 self future fut val 138 3.73 8.04 2.83 4 1
## 19 self past past pref 123 3.96 7.51 2.74 4 2
## 20 self past past pers 123 4.24 8.30 2.88 4 2
## 21 self past past val 123 3.88 9.32 3.05 4 1
## 22 self past fut pref 123 3.77 7.87 2.80 4 1
## 23 self past fut pers 123 3.71 8.55 2.92 3 1
## 24 self past fut val 123 3.34 8.19 2.86 2 1
## q3 min max
## <dbl> <dbl> <dbl>
## 1 6 0 10
## 2 5.75 0 10
## 3 5.75 0 10
## 4 5 0 10
## 5 5 0 10
## 6 5 0 10
## 7 6 0 10
## 8 6 0 10
## 9 5 0 10
## 10 5.5 0 10
## 11 5 0 10
## 12 5 0 10
## 13 6 0 10
## 14 6 0 10
## 15 6 0 10
## 16 6 0 10
## 17 6 0 10
## 18 6 0 10
## 19 6 0 10
## 20 6 0 10
## 21 6 0 10
## 22 6 0 10
## 23 6 0 10
## 24 6 0 10Interpretation:
1. Mean and median values are similar w/ slightly more variation than in
the domain specific anova.
2. Highest to lowest group n size ratio is 1.14 (139/122). Acceptable
ratio for ANOVA (under 1.5).
3. Highest to lowest overall group variance ratio is 1.40 (9.32/6.65).
Acceptable ratio for ANOVA (under 4).
For the sake of consistency w/ the other EHI studies, I also calculated
Hartley’s F-max ratio.
The conservative F-max critical value is 1.60 (same as DS anova since
number of groups and n sizes doesn’t change), which is still higher than
the highest observed F-max ratio of 1.28.
Assumption checks
Missing values
missing_summary <- long_data %>%
group_by(perspective, temporalDO, TIME, DOMAIN) %>%
summarise(
n_total = n(),
n_missing = sum(is.na(DGEN_SCORE)),
pct_missing = round(100 * n_missing / n_total, 2),
.groups = "drop"
)
print(missing_summary, n = Inf)## # A tibble: 24 × 7
## perspective temporalDO TIME DOMAIN n_total n_missing pct_missing
## <fct> <fct> <fct> <fct> <int> <int> <dbl>
## 1 other future past pref 122 0 0
## 2 other future past pers 122 0 0
## 3 other future past val 122 0 0
## 4 other future fut pref 122 0 0
## 5 other future fut pers 122 0 0
## 6 other future fut val 122 0 0
## 7 other past past pref 139 0 0
## 8 other past past pers 139 0 0
## 9 other past past val 139 0 0
## 10 other past fut pref 139 0 0
## 11 other past fut pers 139 0 0
## 12 other past fut val 139 0 0
## 13 self future past pref 138 0 0
## 14 self future past pers 138 0 0
## 15 self future past val 138 0 0
## 16 self future fut pref 138 0 0
## 17 self future fut pers 138 0 0
## 18 self future fut val 138 0 0
## 19 self past past pref 123 0 0
## 20 self past past pers 123 0 0
## 21 self past past val 123 0 0
## 22 self past fut pref 123 0 0
## 23 self past fut pers 123 0 0
## 24 self past fut val 123 0 0No missing values. As expected.
Outliers
outlier_summary <- long_data %>%
group_by(perspective, temporalDO, TIME, DOMAIN) %>%
summarise(
n = n(),
n_outliers = sum(abs(scale(DGEN_SCORE)) > 3),
.groups = "drop"
)
print(outlier_summary, n = Inf)## # A tibble: 24 × 6
## perspective temporalDO TIME DOMAIN n n_outliers
## <fct> <fct> <fct> <fct> <int> <int>
## 1 other future past pref 122 0
## 2 other future past pers 122 0
## 3 other future past val 122 0
## 4 other future fut pref 122 0
## 5 other future fut pers 122 0
## 6 other future fut val 122 0
## 7 other past past pref 139 0
## 8 other past past pers 139 0
## 9 other past past val 139 0
## 10 other past fut pref 139 0
## 11 other past fut pers 139 0
## 12 other past fut val 139 0
## 13 self future past pref 138 0
## 14 self future past pers 138 0
## 15 self future past val 138 0
## 16 self future fut pref 138 0
## 17 self future fut pers 138 0
## 18 self future fut val 138 0
## 19 self past past pref 123 0
## 20 self past past pers 123 0
## 21 self past past val 123 0
## 22 self past fut pref 123 0
## 23 self past fut pers 123 0
## 24 self past fut val 123 0No outliers present. Good. Same as DS anova.
Homogeneity of variance
homogeneity_between <- long_data %>%
group_by(TIME, DOMAIN) %>%
rstatix::levene_test(DGEN_SCORE ~ perspective * temporalDO)
print(homogeneity_between, n = Inf)## # A tibble: 6 × 6
## TIME DOMAIN df1 df2 statistic p
## <fct> <fct> <int> <int> <dbl> <dbl>
## 1 past pref 3 518 0.0902 0.965
## 2 past pers 3 518 0.407 0.748
## 3 past val 3 518 2.69 0.0460
## 4 fut pref 3 518 0.824 0.481
## 5 fut pers 3 518 0.876 0.454
## 6 fut val 3 518 0.238 0.870Levene’s test is signiicant for 1 cell only: past-val. However, variance ratios and F-max tests show that any heteroscedasticity is mild.
Normality (within-subjects residuals)
resid_within <- long_data %>%
group_by(pID) %>%
mutate(person_mean = mean(DGEN_SCORE, na.rm = TRUE)) %>%
ungroup() %>%
mutate(resid = DGEN_SCORE - person_mean) %>%
pull(resid)
resid_within <- resid_within[!is.na(resid_within)]
n_resid <- length(resid_within)
if (n_resid < 3L) {
message("Too few within-subjects residuals (n < 3); skipping Shapiro-Wilk.")
} else {
resid_for_shapiro <- if (n_resid > 5000L) {
set.seed(1L)
sample(resid_within, 5000L)
} else {
resid_within
}
print(shapiro.test(resid_for_shapiro))
}##
## Shapiro-Wilk normality test
##
## data: resid_for_shapiro
## W = 0.94192, p-value < 0.00000000000000022Q-Q plot
qqnorm(resid_within)
qqline(resid_within)
Shapiro-Wilk test is significant but is sensitive to large sample
size.
QQ plot shows that strict centre residuals are normally distributed, but
there is some deviation from normality. ANOVA is robust to violations of
normality w/ large sample size.
Overall, ANOVA can proceed.
Mixed ANOVA
aov_afex <- aov_ez(
id = "pID",
dv = "DGEN_SCORE",
data = long_data,
between = c("perspective", "temporalDO"),
within = c("TIME", "DOMAIN"),
type = 3,
anova_table = list(correction = "none")
)
print(aov_afex)## Anova Table (Type 3 tests)
##
## Response: DGEN_SCORE
## Effect df MSE F ges p.value
## 1 perspective 1, 518 36.26 1.04 .002 .309
## 2 temporalDO 1, 518 36.26 0.03 <.001 .862
## 3 perspective:temporalDO 1, 518 36.26 0.00 <.001 .981
## 4 TIME 1, 518 3.11 8.39 ** .001 .004
## 5 perspective:TIME 1, 518 3.11 0.02 <.001 .890
## 6 temporalDO:TIME 1, 518 3.11 2.94 + <.001 .087
## 7 perspective:temporalDO:TIME 1, 518 3.11 3.44 + <.001 .064
## 8 DOMAIN 2, 1036 2.12 7.85 *** .001 <.001
## 9 perspective:DOMAIN 2, 1036 2.12 1.17 <.001 .312
## 10 temporalDO:DOMAIN 2, 1036 2.12 5.00 ** <.001 .007
## 11 perspective:temporalDO:DOMAIN 2, 1036 2.12 0.39 <.001 .680
## 12 TIME:DOMAIN 2, 1036 1.52 0.77 <.001 .461
## 13 perspective:TIME:DOMAIN 2, 1036 1.52 0.67 <.001 .513
## 14 temporalDO:TIME:DOMAIN 2, 1036 1.52 0.44 <.001 .643
## 15 perspective:temporalDO:TIME:DOMAIN 2, 1036 1.52 3.12 * <.001 .045
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '+' 0.1 ' ' 1Mauchly’s test of sphericity is not significant. Using uncorrected values for interpretation and analysis.
Significant main effects and interactions:
Effect df MSE F ges p 4 TIME 1, 518 3.11 8.39 ** .001 .004
8 DOMAIN 2, 1036 2.13 7.85 *** .001 <.001
10 temporalDO:DOMAIN 2, 1036 2.13 5.00 ** <.001 .007
15 perspective:temporalDO:TIME:DOMAIN 2, 1036 1.52 3.12 * <.001
.045
Mauchly and epsilon
anova_wide <- anova_data %>%
select(pID, perspective, temporalDO, all_of(within_vars)) %>%
filter(if_all(all_of(within_vars), ~ !is.na(.)))
response_matrix <- as.matrix(anova_wide[, within_vars])
rm_model <- lm(response_matrix ~ perspective * temporalDO, data = anova_wide)
idata <- data.frame(
TIME = factor(rep(c("past", "fut"), each = 3), levels = c("past", "fut")),
DOMAIN = factor(rep(c("pref", "pers", "val"), 2), levels = c("pref", "pers", "val"))
)
rm_anova <- car::Anova(rm_model, idata = idata, idesign = ~ TIME * DOMAIN, type = 3)
rm_summary <- summary(rm_anova, multivariate = FALSE)## Warning in summary.Anova.mlm(rm_anova, multivariate = FALSE): HF eps > 1
## treated as 1if (!is.null(rm_summary$sphericity.tests)) {
print(rm_summary$sphericity.tests)
}## Test statistic p-value
## DOMAIN 0.99806 0.60476
## perspective:DOMAIN 0.99806 0.60476
## temporalDO:DOMAIN 0.99806 0.60476
## perspective:temporalDO:DOMAIN 0.99806 0.60476
## TIME:DOMAIN 0.99654 0.40772
## perspective:TIME:DOMAIN 0.99654 0.40772
## temporalDO:TIME:DOMAIN 0.99654 0.40772
## perspective:temporalDO:TIME:DOMAIN 0.99654 0.40772if (!is.null(rm_summary$epsilon)) {
print(rm_summary$epsilon)
}Post hoc comparisons
TIME (main effect)
emm_TIME <- emmeans(aov_afex, ~ TIME)
print(pairs(emm_TIME, adjust = "none"))## contrast estimate SE df t.ratio p.value
## past - fut 0.183 0.0632 518 2.897 0.0039
##
## Results are averaged over the levels of: perspective, temporalDO, DOMAINSupports presence of EHI effect.
DOMAIN (main effect)
emm_DOMAIN <- emmeans(aov_afex, ~ DOMAIN)
print(pairs(emm_DOMAIN, adjust = "tukey"))## contrast estimate SE df t.ratio p.value
## pref - pers 0.129 0.0629 518 2.043 0.1032
## pref - val 0.253 0.0634 518 3.991 0.0002
## pers - val 0.124 0.0652 518 1.908 0.1375
##
## Results are averaged over the levels of: perspective, temporalDO, TIME
## P value adjustment: tukey method for comparing a family of 3 estimatesOnly preference to values contrast is significant.
temporalDO:DOMAIN
emmeans(aov_afex, pairwise ~ temporalDO | DOMAIN, adjust = "none")$contrasts## DOMAIN = pref:
## contrast estimate SE df t.ratio p.value
## future - past -0.00838 0.220 518 -0.038 0.9697
##
## DOMAIN = pers:
## contrast estimate SE df t.ratio p.value
## future - past -0.25252 0.230 518 -1.096 0.2735
##
## DOMAIN = val:
## contrast estimate SE df t.ratio p.value
## future - past 0.14817 0.233 518 0.636 0.5248
##
## Results are averaged over the levels of: perspective, TIMEemmeans(aov_afex, pairwise ~ DOMAIN | temporalDO, adjust = "tukey")$contrasts## temporalDO = future:
## contrast estimate SE df t.ratio p.value
## pref - pers 0.25065 0.0892 518 2.810 0.0142
## pref - val 0.17474 0.0898 518 1.945 0.1271
## pers - val -0.07591 0.0924 518 -0.821 0.6900
##
## temporalDO = past:
## contrast estimate SE df t.ratio p.value
## pref - pers 0.00651 0.0888 518 0.073 0.9970
## pref - val 0.33129 0.0895 518 3.702 0.0007
## pers - val 0.32478 0.0921 518 3.527 0.0013
##
## Results are averaged over the levels of: perspective, TIME
## P value adjustment: tukey method for comparing a family of 3 estimatesWhen grouped by domain, no contrasts are significant.
When grouped by temporalDO, some contrasts are significant:
Future-first temporal display order:
contrast estimate SE df t.ratio p.value
pref - pers 0.25065 0.0892 518 2.810 0.0142
Past-first temporal display order:
contrast estimate SE df t.ratio p.value
pref - val 0.33129 0.0895 518 3.702 0.0007
pers - val 0.32478 0.0921 518 3.527 0.0013
perspective:temporalDO:TIME:DOMAIN
contrasts for TIME grouped by perspective, temporalDO, and DOMAIN
emm_fourway <- emmeans(aov_afex, pairwise ~ TIME | perspective * temporalDO * DOMAIN, adjust = "tukey")
print(emm_fourway$contrasts)## perspective = other, temporalDO = future, DOMAIN = pref:
## contrast estimate SE df t.ratio p.value
## past - fut 0.0492 0.182 518 0.270 0.7873
##
## perspective = self, temporalDO = future, DOMAIN = pref:
## contrast estimate SE df t.ratio p.value
## past - fut -0.0362 0.171 518 -0.212 0.8326
##
## perspective = other, temporalDO = past, DOMAIN = pref:
## contrast estimate SE df t.ratio p.value
## past - fut 0.3237 0.171 518 1.897 0.0584
##
## perspective = self, temporalDO = past, DOMAIN = pref:
## contrast estimate SE df t.ratio p.value
## past - fut 0.1870 0.181 518 1.031 0.3032
##
## perspective = other, temporalDO = future, DOMAIN = pers:
## contrast estimate SE df t.ratio p.value
## past - fut 0.2049 0.179 518 1.142 0.2540
##
## perspective = self, temporalDO = future, DOMAIN = pers:
## contrast estimate SE df t.ratio p.value
## past - fut 0.0145 0.169 518 0.086 0.9316
##
## perspective = other, temporalDO = past, DOMAIN = pers:
## contrast estimate SE df t.ratio p.value
## past - fut 0.2878 0.168 518 1.712 0.0875
##
## perspective = self, temporalDO = past, DOMAIN = pers:
## contrast estimate SE df t.ratio p.value
## past - fut 0.5285 0.179 518 2.957 0.0032
##
## perspective = other, temporalDO = future, DOMAIN = val:
## contrast estimate SE df t.ratio p.value
## past - fut 0.2951 0.188 518 1.568 0.1175
##
## perspective = self, temporalDO = future, DOMAIN = val:
## contrast estimate SE df t.ratio p.value
## past - fut -0.0797 0.177 518 -0.450 0.6526
##
## perspective = other, temporalDO = past, DOMAIN = val:
## contrast estimate SE df t.ratio p.value
## past - fut -0.1151 0.176 518 -0.653 0.5141
##
## perspective = self, temporalDO = past, DOMAIN = val:
## contrast estimate SE df t.ratio p.value
## past - fut 0.5366 0.187 518 2.863 0.0044Significant contrasts:
contrast estimate SE df t.ratio p.value
past - fut 0.5285 0.179 518 2.957 0.0032 (self-perspective, personality
domain, past-first temporal display order)
past - fut 0.5366 0.187 518 2.863 0.0044 (self-perspective, values
domain, past-first temporal display order)
### contrasts for DOMAIN grouped by perspective, TIME, and temporalDO
emm_fourway2 <- emmeans(aov_afex, pairwise ~ DOMAIN | perspective * TIME * temporalDO, adjust = "tukey")
print(emm_fourway2$contrasts)## perspective = other, TIME = past, temporalDO = future:
## contrast estimate SE df t.ratio p.value
## pref - pers 0.2459 0.168 518 1.468 0.3073
## pref - val 0.0164 0.177 518 0.093 0.9953
## pers - val -0.2295 0.171 518 -1.343 0.3718
##
## perspective = self, TIME = past, temporalDO = future:
## contrast estimate SE df t.ratio p.value
## pref - pers 0.1522 0.158 518 0.966 0.5986
## pref - val 0.2319 0.166 518 1.395 0.3445
## pers - val 0.0797 0.161 518 0.496 0.8732
##
## perspective = other, TIME = fut, temporalDO = future:
## contrast estimate SE df t.ratio p.value
## pref - pers 0.4016 0.177 518 2.268 0.0613
## pref - val 0.2623 0.169 518 1.552 0.2678
## pers - val -0.1393 0.175 518 -0.798 0.7046
##
## perspective = self, TIME = fut, temporalDO = future:
## contrast estimate SE df t.ratio p.value
## pref - pers 0.2029 0.167 518 1.219 0.4427
## pref - val 0.1884 0.159 518 1.185 0.4625
## pers - val -0.0145 0.164 518 -0.088 0.9957
##
## perspective = other, TIME = past, temporalDO = past:
## contrast estimate SE df t.ratio p.value
## pref - pers 0.1367 0.157 518 0.871 0.6589
## pref - val 0.6259 0.166 518 3.778 0.0005
## pers - val 0.4892 0.160 518 3.056 0.0066
##
## perspective = self, TIME = past, temporalDO = past:
## contrast estimate SE df t.ratio p.value
## pref - pers -0.2764 0.167 518 -1.657 0.2230
## pref - val 0.0813 0.176 518 0.462 0.8892
## pers - val 0.3577 0.170 518 2.102 0.0903
##
## perspective = other, TIME = fut, temporalDO = past:
## contrast estimate SE df t.ratio p.value
## pref - pers 0.1007 0.166 518 0.607 0.8163
## pref - val 0.1871 0.158 518 1.181 0.4650
## pers - val 0.0863 0.164 518 0.528 0.8579
##
## perspective = self, TIME = fut, temporalDO = past:
## contrast estimate SE df t.ratio p.value
## pref - pers 0.0650 0.176 518 0.369 0.9278
## pref - val 0.4309 0.168 518 2.559 0.0290
## pers - val 0.3659 0.174 518 2.103 0.0902
##
## P value adjustment: tukey method for comparing a family of 3 estimatesSignificant contrasts:
contrast estimate SE df t.ratio p.value
pref - val 0.6259 0.166 518 3.778 0.0005 (other-perspective,
past-directed questions, past-first temporal display order)
pers - val 0.4892 0.160 518 3.056 0.0066 (other-perspective,
past-directed questions, past-first temporal display order)
pref - val 0.4309 0.168 518 2.559 0.0290 (self-perspective,
future-directed questions, past-first temporal display order)
Cohen’s d (significant contrasts only)
d_data <- anova_data %>%
mutate(
past_mean = (past_pref_DGEN + past_pers_DGEN + past_val_DGEN) / 3,
fut_mean = (fut_pref_DGEN + fut_pers_DGEN + fut_val_DGEN) / 3,
pref_mean = (past_pref_DGEN + fut_pref_DGEN) / 2,
pers_mean = (past_pers_DGEN + fut_pers_DGEN) / 2,
val_mean = (past_val_DGEN + fut_val_DGEN) / 2
)
cohens_d_results <- tibble(
contrast = character(),
condition = character(),
d = double()
)
# TIME main: past vs fut
cohens_d_results <- cohens_d_results %>%
add_row(
contrast = "TIME (past - fut)",
condition = "overall",
d = suppressMessages(effectsize::cohens_d(d_data$past_mean, d_data$fut_mean, paired = TRUE)$Cohens_d)
)
# DOMAIN main: pref vs val
cohens_d_results <- cohens_d_results %>%
add_row(
contrast = "DOMAIN (pref - val)",
condition = "overall",
d = suppressMessages(effectsize::cohens_d(d_data$pref_mean, d_data$val_mean, paired = TRUE)$Cohens_d)
)
# temporalDO:DOMAIN - future: pref vs pers
d_fut <- d_data %>% filter(temporalDO == "future")
cohens_d_results <- cohens_d_results %>%
add_row(
contrast = "DOMAIN (pref - pers)",
condition = "temporalDO = future",
d = suppressMessages(effectsize::cohens_d(d_fut$pref_mean, d_fut$pers_mean, paired = TRUE)$Cohens_d)
)
# temporalDO:DOMAIN - past: pref vs val, pers vs val
d_past <- d_data %>% filter(temporalDO == "past")
cohens_d_results <- cohens_d_results %>%
add_row(
contrast = "DOMAIN (pref - val)",
condition = "temporalDO = past",
d = suppressMessages(effectsize::cohens_d(d_past$pref_mean, d_past$val_mean, paired = TRUE)$Cohens_d)
) %>%
add_row(
contrast = "DOMAIN (pers - val)",
condition = "temporalDO = past",
d = suppressMessages(effectsize::cohens_d(d_past$pers_mean, d_past$val_mean, paired = TRUE)$Cohens_d)
)
# 4-way TIME: self, past temporalDO, pers
d_self_past <- d_data %>% filter(perspective == "self", temporalDO == "past")
cohens_d_results <- cohens_d_results %>%
add_row(
contrast = "TIME (past - fut)",
condition = "self, past temporalDO, pers domain",
d = suppressMessages(effectsize::cohens_d(d_self_past$past_pers_DGEN, d_self_past$fut_pers_DGEN, paired = TRUE)$Cohens_d)
)
# 4-way TIME: self, past temporalDO, val
cohens_d_results <- cohens_d_results %>%
add_row(
contrast = "TIME (past - fut)",
condition = "self, past temporalDO, val domain",
d = suppressMessages(effectsize::cohens_d(d_self_past$past_val_DGEN, d_self_past$fut_val_DGEN, paired = TRUE)$Cohens_d)
)
# 4-way DOMAIN: other, past TIME, past temporalDO - pref vs val
d_other_past_tpast <- d_data %>% filter(perspective == "other", temporalDO == "past")
cohens_d_results <- cohens_d_results %>%
add_row(
contrast = "DOMAIN (pref - val)",
condition = "other, past TIME, past temporalDO",
d = suppressMessages(effectsize::cohens_d(d_other_past_tpast$past_pref_DGEN, d_other_past_tpast$past_val_DGEN, paired = TRUE)$Cohens_d)
)
# 4-way DOMAIN: other, past TIME, past temporalDO - pers vs val
cohens_d_results <- cohens_d_results %>%
add_row(
contrast = "DOMAIN (pers - val)",
condition = "other, past TIME, past temporalDO",
d = suppressMessages(effectsize::cohens_d(d_other_past_tpast$past_pers_DGEN, d_other_past_tpast$past_val_DGEN, paired = TRUE)$Cohens_d)
)
# 4-way DOMAIN: self, fut TIME, past temporalDO - pref vs val
d_self_fut_tpast <- d_data %>% filter(perspective == "self", temporalDO == "past")
cohens_d_results <- cohens_d_results %>%
add_row(
contrast = "DOMAIN (pref - val)",
condition = "self, fut TIME, past temporalDO",
d = suppressMessages(effectsize::cohens_d(d_self_fut_tpast$fut_pref_DGEN, d_self_fut_tpast$fut_val_DGEN, paired = TRUE)$Cohens_d)
)
cohens_d_results %>%
mutate(d = round(d, 3)) %>%
print(n = Inf)## # A tibble: 10 × 3
## contrast condition d
## <chr> <chr> <dbl>
## 1 TIME (past - fut) overall 0.122
## 2 DOMAIN (pref - val) overall 0.178
## 3 DOMAIN (pref - pers) temporalDO = future 0.174
## 4 DOMAIN (pref - val) temporalDO = past 0.242
## 5 DOMAIN (pers - val) temporalDO = past 0.239
## 6 TIME (past - fut) self, past temporalDO, pers domain 0.26
## 7 TIME (past - fut) self, past temporalDO, val domain 0.254
## 8 DOMAIN (pref - val) other, past TIME, past temporalDO 0.348
## 9 DOMAIN (pers - val) other, past TIME, past temporalDO 0.241
## 10 DOMAIN (pref - val) self, fut TIME, past temporalDO 0.257Size d Interpretation
Small 0.2 Weak effect
Medium 0.5 Moderate effect
Large 0.8 Strong effect