Gesture and Language Meta-Analysis
2021-07-29
Here are the analyses without Spearman correlation’s coefficients. For analyses on the full data set click here.
See R Markdown file for code used to open data set and calculated effect size variances etc. It prints kind of messy, and I have suppressed it from this HTML file.
1 Determining Baseline Model
Fit baseline model to the full data set
m1 <- rma.mv(yi, vi, random=~1|Article/Comp, data=EffectSizes_short)
summary(m1)##
## Multivariate Meta-Analysis Model (k = 87; method: REML)
##
## logLik Deviance AIC BIC AICc
## -5.5354 11.0708 17.0708 24.4338 17.3634
##
## Variance Components:
##
## estim sqrt nlvls fixed factor
## sigma^2.1 0.0381 0.1952 27 no Article
## sigma^2.2 0.0027 0.0518 87 no Article/Comp
##
## Test for Heterogeneity:
## Q(df = 86) = 206.7883, p-val < .0001
##
## Model Results:
##
## estimate se zval pval ci.lb ci.ub
## 0.2572 0.0461 5.5758 <.0001 0.1668 0.3476 ***
##
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1Look at residuals
boxplot(resid(m1, method="rstandard"))
hist(resid(m1, method="rstandard"))
These residuals aren’t great.
Summary, controlling for correlated sampling error.
summary(robust(m1, cluster=EffectSizes_short$Article))##
## Number of outcomes: 87
## Number of clusters: 27
## Outcomes per cluster: 1-8 (mean: 3.22, median: 2)
##
## Model Results:
##
## estimate se tval df pval ci.lb ci.ub
## 0.2572 0.0460 5.5859 26 <.0001 0.1625 0.3518 ***
##
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1Tests of publication bias.
regtest(rma(yi, vi, data=EffectSizes_short))##
## Regression Test for Funnel Plot Asymmetry
##
## Model: mixed-effects meta-regression model
## Predictor: standard error
##
## Test for Funnel Plot Asymmetry: z = 1.4447, p = 0.1486
## Limit Estimate (as sei -> 0): b = 0.1430 (CI: 0.0433, 0.2428)funnel(rma(yi, vi, data=EffectSizes_short))
There are a lot of effect size variances that are quite large because a subset of the correlaions were calculated on very small sample sizes. These variances could potentially strongly influence the slope of the corrections for effect sizes. If we look at a histogram of the effect size standard errors, we see that all of the especially large effect sizes standard errors come from studies with sample sizes < = 10.
EffectSizes_short %>%
mutate(
small_n = ifelse(n <= 10, yes="N <= 10", no = "N > 10")
) %>%
ggplot(aes(x = sei, fill=small_n)) + geom_histogram()## `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.
Given that, these effect sizes could strongly influence the publication bias correction, and that effect sizes based on 10 participants are likely not very informative to begin with. I have decided to drop these corrections from subsequent analyses.
Here I fit the model again, without those effect sizes.
m2 <- rma.mv(yi, vi, random=~1|Article/Comp, data=EffectSizes_short2)
summary(m2)##
## Multivariate Meta-Analysis Model (k = 71; method: REML)
##
## logLik Deviance AIC BIC AICc
## 15.5986 -31.1972 -25.1972 -18.4518 -24.8336
##
## Variance Components:
##
## estim sqrt nlvls fixed factor
## sigma^2.1 0.0354 0.1880 23 no Article
## sigma^2.2 0.0026 0.0512 71 no Article/Comp
##
## Test for Heterogeneity:
## Q(df = 70) = 175.5400, p-val < .0001
##
## Model Results:
##
## estimate se zval pval ci.lb ci.ub
## 0.2546 0.0464 5.4876 <.0001 0.1636 0.3455 ***
##
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1and test for publication bias again
regtest(rma(yi, vi, data=EffectSizes_short2))##
## Regression Test for Funnel Plot Asymmetry
##
## Model: mixed-effects meta-regression model
## Predictor: standard error
##
## Test for Funnel Plot Asymmetry: z = 2.5729, p = 0.0101
## Limit Estimate (as sei -> 0): b = 0.0567 (CI: -0.0650, 0.1784)funnel(rma(yi, vi, data=EffectSizes_short2))
The funnel plots and regression test still suggest publication bias, but there aren’t the disproporionately large effect sizes that would influence the publication bias correction.
boxplot(resid(m2, method="rstandard"))
hist(resid(m2, method="rstandard"))
The residuals look much better, if still a bit skewed.
Here’s a forest plot
forest(m2, slab=EffectSizes_short2$Article, xlim=c(-2.8,2.8))
Add variance to correct for publication bias
m3 <- rma.mv(yi, vi, mods=vi, random=~1|Article/Comp, data=EffectSizes_short2)
summary(m3)##
## Multivariate Meta-Analysis Model (k = 71; method: REML)
##
## logLik Deviance AIC BIC AICc
## 15.7316 -31.4632 -23.4632 -14.5268 -22.8382
##
## Variance Components:
##
## estim sqrt nlvls fixed factor
## sigma^2.1 0.0372 0.1928 23 no Article
## sigma^2.2 0.0026 0.0512 71 no Article/Comp
##
## Test for Residual Heterogeneity:
## QE(df = 69) = 150.0938, p-val < .0001
##
## Test of Moderators (coefficient 2):
## QM(df = 1) = 0.0695, p-val = 0.7921
##
## Model Results:
##
## estimate se zval pval ci.lb ci.ub
## intrcpt 0.2367 0.0841 2.8153 0.0049 0.0719 0.4015 **
## mods 0.5583 2.1183 0.2635 0.7921 -3.5936 4.7101
##
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1m3 <- rma.mv(yi, vi, mods=~vi, random=~1|Article/Comp, data=EffectSizes_short2)
summary(m3)##
## Multivariate Meta-Analysis Model (k = 71; method: REML)
##
## logLik Deviance AIC BIC AICc
## 15.7316 -31.4632 -23.4632 -14.5268 -22.8382
##
## Variance Components:
##
## estim sqrt nlvls fixed factor
## sigma^2.1 0.0372 0.1928 23 no Article
## sigma^2.2 0.0026 0.0512 71 no Article/Comp
##
## Test for Residual Heterogeneity:
## QE(df = 69) = 150.0938, p-val < .0001
##
## Test of Moderators (coefficient 2):
## QM(df = 1) = 0.0695, p-val = 0.7921
##
## Model Results:
##
## estimate se zval pval ci.lb ci.ub
## intrcpt 0.2367 0.0841 2.8153 0.0049 0.0719 0.4015 **
## vi 0.5583 2.1183 0.2635 0.7921 -3.5936 4.7101
##
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1boxplot(resid(m3, method="rstandard"))
hist(resid(m3, method="rstandard"))
Summarize the above model, with cluster robust variance estimation, by article, to account for the correlated sampling errors within articles.
summary(robust(m3, cluster=EffectSizes_short2$Article))##
## Number of outcomes: 71
## Number of clusters: 23
## Outcomes per cluster: 1-8 (mean: 3.09, median: 2)
##
## Test of Moderators (coefficient 2):
## F(df1 = 1, df2 = 21) = 0.1060, p-val = 0.7479
##
## Model Results:
##
## estimate se tval df pval ci.lb ci.ub
## intrcpt 0.2367 0.0828 2.8578 21 0.0094 0.0645 0.4090 **
## vi 0.5583 1.7144 0.3256 21 0.7479 -3.0069 4.1235
##
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 12 Tests of the Moderators
Here are all of the analyses of the moderator variables. The code for each moderator is basically the same, so I have annotated it for the model testing the effect of modality only.
2.1 Modality: Comprehension or Production
m_modal <- rma.mv(yi, vi, mods= ~ Modalitys*vi, random=~1|Article/Comp, data=EffectSizes_short2) # fit multilevel model.
summary(robust(m_modal, cluster=EffectSizes_short2$Article)) # get robust standard errors##
## Number of outcomes: 71
## Number of clusters: 23
## Outcomes per cluster: 1-8 (mean: 3.09, median: 2)
##
## Test of Moderators (coefficients 2:4):
## F(df1 = 3, df2 = 19) = 18.3082, p-val < .0001
##
## Model Results:
##
## estimate se tval df pval ci.lb ci.ub
## intrcpt 0.2512 0.0813 3.0883 19 0.0061 0.0809 0.4214 **
## Modalitys1 0.0704 0.0139 5.0752 19 <.0001 0.0414 0.0994 ***
## vi 0.0354 1.5031 0.0236 19 0.9814 -3.1107 3.1815
## Modalitys1:vi -2.0670 0.5903 -3.5013 19 0.0024 -3.3026 -0.8314 **
##
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1anova(robust(m_modal, cluster=EffectSizes_short2$Article), btt=2) # F test with robust standard errors##
## Test of Moderators (coefficient 2):
## F(df1 = 1, df2 = 19) = 25.7578, p-val < .0001predict(robust(m_modal, cluster=EffectSizes_short2$Article), c(1, 0, 0), transf=transf.ztor) # get means and CIs for level 1 (assuming 0 se)##
## pred ci.lb ci.ub pi.lb pi.ub
## 0.3109 0.1624 0.4457 -0.1099 0.6372predict(robust(m_modal, cluster=EffectSizes_short2$Article), c(-1, 0, 0), transf=transf.ztor) # get means and CIs for level -1 (assuming 0 se)##
## pred ci.lb ci.ub pi.lb pi.ub
## 0.1789 -0.0056 0.3516 -0.2566 0.5539table(EffectSizes_short2$Modalitys)##
## comp prod
## 27 44contrasts(EffectSizes_short2$Modalitys) # Just a sanity check: which is -1 and which is 1## [,1]
## comp 1
## prod -1EffectSizes_short2 %>% # get descriptive statistics for each level of the variable.
group_by(Modalitys) %>%
summarise(
N_part = sum(n),
K = n()
)2.2 Measure: Vocab vs Other
m_measure <- rma.mv(yi, vi, mods= ~ Measures*vi, random=~1|Article/Comp, data=EffectSizes_short2)## Warning: Rows with NAs omitted from model fitting.summary(robust(m_measure, cluster=EffectSizes_short2$Article))##
## Number of outcomes: 69
## Number of clusters: 23
## Outcomes per cluster: 1-8 (mean: 3.00, median: 2)
##
## Test of Moderators (coefficients 2:4):
## F(df1 = 3, df2 = 19) = 2.6858, p-val = 0.0757
##
## Model Results:
##
## estimate se tval df pval ci.lb ci.ub
## intrcpt 0.2366 0.0893 2.6503 19 0.0158 0.0497 0.4234 *
## Measures1 0.0282 0.0118 2.3906 19 0.0273 0.0035 0.0529 *
## vi 0.8845 1.8015 0.4910 19 0.6291 -2.8860 4.6550
## Measures1:vi 0.2222 0.7488 0.2967 19 0.7699 -1.3451 1.7895
##
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1anova(robust(m_measure, cluster=EffectSizes_short2$Article), btt=2)##
## Test of Moderators (coefficient 2):
## F(df1 = 1, df2 = 19) = 5.7149, p-val = 0.0273predict(robust(m_measure, cluster=EffectSizes_short2$Article), c(1, 0, 0), transf=transf.ztor)##
## pred ci.lb ci.ub pi.lb pi.ub
## 0.2588 0.0829 0.4189 -0.1944 0.6209predict(robust(m_measure, cluster=EffectSizes_short2$Article), c(-1, 0, 0), transf=transf.ztor)##
## pred ci.lb ci.ub pi.lb pi.ub
## 0.2054 0.0133 0.3829 -0.2531 0.5886table(EffectSizes_short2$Measures)##
## Other Vocab
## 27 42contrasts(EffectSizes_short2$Measures)## [,1]
## Other 1
## Vocab -1EffectSizes_short2 %>%
group_by(Measures) %>%
summarise(
N_part = sum(n),
K = n()
)2.3 Pointing Context: Gesture Alone vs Gesture and Speech Combination
m_context <- rma.mv(yi, vi, mods= ~ Just_Pointings*vi, random=~1|Article/Comp, data=EffectSizes_short2)
summary(robust(m_context, cluster=EffectSizes_short2$Article))##
## Number of outcomes: 71
## Number of clusters: 23
## Outcomes per cluster: 1-8 (mean: 3.09, median: 2)
##
## Test of Moderators (coefficients 2:4):
## F(df1 = 3, df2 = 19) = 12.4862, p-val < .0001
##
## Model Results:
##
## estimate se tval df pval ci.lb ci.ub
## intrcpt 0.1136 0.0839 1.3531 19 0.1919 -0.0621 0.2892
## Just_Pointings1 0.1347 0.0305 4.4197 19 0.0003 0.0709 0.1985
## vi 4.8324 1.7441 2.7707 19 0.0122 1.1820 8.4828
## Just_Pointings1:vi -4.9678 1.4378 -3.4551 19 0.0027 -7.9772 -1.9584
##
## intrcpt
## Just_Pointings1 ***
## vi *
## Just_Pointings1:vi **
##
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1anova(robust(m_context, cluster=EffectSizes_short2$Article), btt=2)##
## Test of Moderators (coefficient 2):
## F(df1 = 1, df2 = 19) = 19.5342, p-val = 0.0003table(EffectSizes_short2$Just_Pointings)##
## pointing_alone Gesture_Speech_Combination
## 64 7predict(robust(m_context, cluster=EffectSizes_short2$Article), c(1, 0, 0), transf=transf.ztor)##
## pred ci.lb ci.ub pi.lb pi.ub
## 0.2433 0.0628 0.4083 -0.2230 0.6190predict(robust(m_context, cluster=EffectSizes_short2$Article), c(-1, 0, 0), transf=transf.ztor)##
## pred ci.lb ci.ub pi.lb pi.ub
## -0.0211 -0.2065 0.1657 -0.4601 0.4261table(EffectSizes_short2$Just_Pointings)##
## pointing_alone Gesture_Speech_Combination
## 64 7EffectSizes_short2 %>%
group_by(Just_Pointings) %>%
summarise(
N_part = sum(n),
K = n()
)2.4 Age of Pointing. < 18 month or > 18 month.
m_pointing <- rma.mv(yi, vi, mods= ~ Age_Pointings*vi , random=~1|Article/Comp, data=EffectSizes_short2)
summary(robust(m_pointing, cluster=EffectSizes_short2$Article))##
## Number of outcomes: 71
## Number of clusters: 23
## Outcomes per cluster: 1-8 (mean: 3.09, median: 2)
##
## Test of Moderators (coefficients 2:4):
## F(df1 = 3, df2 = 19) = 375.9006, p-val < .0001
##
## Model Results:
##
## estimate se tval df pval ci.lb ci.ub
## intrcpt 0.2901 0.0939 3.0897 19 0.0060 0.0936 0.4866 **
## Age_Pointings1 0.0604 0.0056 10.8430 19 <.0001 0.0487 0.0720 ***
## vi 0.2348 1.9025 0.1234 19 0.9031 -3.7472 4.2169
## Age_Pointings1:vi -0.0157 0.7042 -0.0223 19 0.9824 -1.4895 1.4581
##
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1anova(robust(m_pointing, cluster=EffectSizes_short2$Article), btt=2)##
## Test of Moderators (coefficient 2):
## F(df1 = 1, df2 = 19) = 117.5698, p-val < .0001table(EffectSizes_short2$Age_Pointings)##
## Over_18 Under 18
## 11 60predict(robust(m_pointing, cluster=EffectSizes_short2$Article), c(1, 0, 0), transf=transf.ztor)##
## pred ci.lb ci.ub pi.lb pi.ub
## 0.3368 0.1622 0.4909 -0.1242 0.6782predict(robust(m_pointing, cluster=EffectSizes_short2$Article), c(-1, 0, 0), transf=transf.ztor)##
## pred ci.lb ci.ub pi.lb pi.ub
## 0.2258 0.0233 0.4104 -0.2483 0.6126EffectSizes_short2 %>%
group_by(Age_Pointings) %>%
summarise(
N_part = sum(n),
K = n()
)2.5 Language: English, Mixed, Other
m_language <- rma.mv(yi, vi, mods= ~ Languages*vi, random=~1|Article/Comp, data=EffectSizes_short2)## Warning: Redundant predictors dropped from the model.summary(robust(m_language, cluster=EffectSizes_short2$Article))##
## Number of outcomes: 71
## Number of clusters: 23
## Outcomes per cluster: 1-8 (mean: 3.09, median: 2)
##
## Test of Moderators (coefficients 2:5):
## F(df1 = 4, df2 = 18) = 12.0935, p-val < .0001
##
## Model Results:
##
## estimate se tval df pval ci.lb ci.ub
## intrcpt 0.1263 0.0614 2.0575 18 0.0544 -0.0027 0.2552 .
## Languages1 0.2100 0.0895 2.3458 18 0.0306 0.0219 0.3981 *
## Languages2 -0.1059 0.0547 -1.9376 18 0.0685 -0.2207 0.0089 .
## vi 1.6932 1.7269 0.9805 18 0.3398 -1.9349 5.3213
## Languages1:vi -2.2328 1.7269 -1.2929 18 0.2124 -5.8609 1.3954
##
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1anova(robust(m_language, cluster=EffectSizes_short2$Article), btt=c(2, 3))##
## Test of Moderators (coefficients 2:3):
## F(df1 = 2, df2 = 18) = 3.9237, p-val = 0.0385table(EffectSizes_short2$Languages)##
## English Mixed Other
## 31 8 32predict(robust(m_language, cluster=EffectSizes_short2$Article), c(-1, -1, 0, 0), transf=transf.ztor) #Other##
## pred ci.lb ci.ub pi.lb pi.ub
## 0.0221 -0.2511 0.2921 -0.4416 0.4765predict(robust(m_language, cluster=EffectSizes_short2$Article), c(1, 0, 0, 0), transf=transf.ztor) #English##
## pred ci.lb ci.ub pi.lb pi.ub
## 0.3241 0.0889 0.5251 -0.1420 0.6726predict(robust(m_language, cluster=EffectSizes_short2$Article), c(0, 1, 0, 0), transf=transf.ztor) #Mixed##
## pred ci.lb ci.ub pi.lb pi.ub
## 0.0204 0.0041 0.0366 -0.3719 0.4065contrasts(EffectSizes_short2$Languages)## [,1] [,2]
## English 1 0
## Mixed 0 1
## Other -1 -1EffectSizes_short2 %>%
group_by(Languages) %>%
summarise(
N_part = sum(n),
K = n(),
mean = median(yi)
)2.6 Country: North America/Other
m_country <- rma.mv(yi, vi, mods= ~ Countrys*vi , random=~1|Article/Comp, data=EffectSizes_short2)## Warning: Redundant predictors dropped from the model.summary(robust(m_country, cluster=EffectSizes_short2$Article))##
## Number of outcomes: 71
## Number of clusters: 23
## Outcomes per cluster: 1-8 (mean: 3.09, median: 2)
##
## Test of Moderators (coefficients 2:4):
## F(df1 = 3, df2 = 19) = 1.3935, p-val = 0.2754
##
## Model Results:
##
## estimate se tval df pval ci.lb ci.ub
## intrcpt 0.1901 0.0878 2.1666 19 0.0432 0.0065 0.3738 *
## Countrys1 0.1360 0.0878 1.5499 19 0.1377 -0.0477 0.3197
## vi 1.7230 1.8980 0.9078 19 0.3754 -2.2496 5.6955
## Countrys1:vi -1.4138 1.8980 -0.7449 19 0.4655 -5.3863 2.5587
##
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1anova(robust(m_country, cluster=EffectSizes_short2$Article), btt=2)##
## Test of Moderators (coefficient 2):
## F(df1 = 1, df2 = 19) = 2.4021, p-val = 0.1377table(EffectSizes_short2$Countrys)##
## North America Mixed Other
## 34 0 37predict(robust(m_country, cluster=EffectSizes_short2$Article), c(1, 0, 0), transf=transf.ztor)##
## pred ci.lb ci.ub pi.lb pi.ub
## 0.3151 0.0598 0.5317 -0.1534 0.6679predict(robust(m_country, cluster=EffectSizes_short2$Article), c(-1, 0, 0), transf=transf.ztor)##
## pred ci.lb ci.ub pi.lb pi.ub
## 0.0541 -0.1963 0.2979 -0.3965 0.4836EffectSizes_short2 %>%
group_by(Countrys) %>%
summarise(
N_part = sum(n),
K = n()
)2.7 Pointing Measure: Frequency/Onset
m_fo <- rma.mv(yi, vi, mods= ~ Pointing_Measures + vi, random=~1|Article/Comp, data=EffectSizes_short2)
summary(robust(m_fo, cluster=EffectSizes_short2$Article))##
## Number of outcomes: 71
## Number of clusters: 23
## Outcomes per cluster: 1-8 (mean: 3.09, median: 2)
##
## Test of Moderators (coefficients 2:3):
## F(df1 = 2, df2 = 20) = 0.0482, p-val = 0.9530
##
## Model Results:
##
## estimate se tval df pval ci.lb ci.ub
## intrcpt 0.2342 0.1115 2.0999 20 0.0486 0.0015 0.4669 *
## Pointing_Measures1 0.0040 0.0696 0.0570 20 0.9551 -0.1411 0.1491
## vi 0.5778 1.9039 0.3035 20 0.7647 -3.3936 4.5492
##
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1table(EffectSizes_short2$Pointing_Measures)##
## frequency onset
## 58 13anova(robust(m_fo, cluster=EffectSizes_short2$Article), btt=2)##
## Test of Moderators (coefficient 2):
## F(df1 = 1, df2 = 20) = 0.0032, p-val = 0.9551predict(robust(m_fo, cluster=EffectSizes_short2$Article), c(1, 0), transf=transf.ztor)##
## pred ci.lb ci.ub pi.lb pi.ub
## 0.2338 0.0631 0.3912 -0.2208 0.6049predict(robust(m_fo, cluster=EffectSizes_short2$Article), c(-1, 0), transf=transf.ztor)##
## pred ci.lb ci.ub pi.lb pi.ub
## 0.2263 -0.1153 0.5200 -0.3099 0.6532contrasts(EffectSizes_short2$Pointing_Measures)## [,1]
## frequency 1
## onset -1EffectSizes_short2 %>%
group_by(Pointing_Measures) %>%
summarise(
N_part = sum(n),
K = n(),
S = sum(as.integer(Pointing_Measures)),
mean(yi)
)# I got strange predicted values when I allowed the interaction between variance and condition, so I removed the interaction. 2.8 Design: Concurrent/Longitudinal
m_design <- rma.mv(yi, vi, mods= ~ Designs*vi, random=~1|Article/Comp, data=EffectSizes_short2)
summary(robust(m_design, cluster=EffectSizes_short2$Article))##
## Number of outcomes: 71
## Number of clusters: 23
## Outcomes per cluster: 1-8 (mean: 3.09, median: 2)
##
## Test of Moderators (coefficients 2:4):
## F(df1 = 3, df2 = 19) = 1.5816, p-val = 0.2268
##
## Model Results:
##
## estimate se tval df pval ci.lb ci.ub
## intrcpt 0.2420 0.0888 2.7240 19 0.0135 0.0561 0.4279 *
## Designs1 0.0300 0.0212 1.4109 19 0.1745 -0.0145 0.0744
## vi 0.4948 1.6133 0.3067 19 0.7624 -2.8819 3.8716
## Designs1:vi -0.8769 0.8958 -0.9789 19 0.3399 -2.7519 0.9980
##
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1table(EffectSizes_short2$Designs)##
## Concurrent Longitudinal
## 18 53anova(robust(m_design, cluster=EffectSizes_short2$Article), btt=2)##
## Test of Moderators (coefficient 2):
## F(df1 = 1, df2 = 19) = 1.9905, p-val = 0.1745predict(robust(m_design, cluster=EffectSizes_short2$Article), c(1, 0, 0), transf=transf.ztor)##
## pred ci.lb ci.ub pi.lb pi.ub
## 0.2654 0.0560 0.4525 -0.1940 0.6294predict(robust(m_design, cluster=EffectSizes_short2$Article), c(-1, 0, 0), transf=transf.ztor)##
## pred ci.lb ci.ub pi.lb pi.ub
## 0.2089 0.0492 0.3582 -0.2303 0.5774contrasts(EffectSizes_short2$Designs)## [,1]
## Concurrent 1
## Longitudinal -1EffectSizes_short2 %>%
group_by(Designs) %>%
summarise(
N_part = sum(n),
K = n()
)sessionInfo()## R version 4.0.2 (2020-06-22)
## Platform: x86_64-w64-mingw32/x64 (64-bit)
## Running under: Windows 10 x64 (build 19043)
##
## Matrix products: default
##
## locale:
## [1] LC_COLLATE=Dutch_Netherlands.1252 LC_CTYPE=Dutch_Netherlands.1252
## [3] LC_MONETARY=Dutch_Netherlands.1252 LC_NUMERIC=C
## [5] LC_TIME=Dutch_Netherlands.1252
##
## attached base packages:
## [1] stats graphics grDevices utils datasets methods base
##
## other attached packages:
## [1] forcats_0.5.1 stringr_1.4.0 dplyr_1.0.5 purrr_0.3.4
## [5] tidyr_1.1.3 tibble_3.1.1 ggplot2_3.3.3 tidyverse_1.3.1
## [9] metafor_3.0-2 Matrix_1.2-18 readr_1.4.0
##
## loaded via a namespace (and not attached):
## [1] tidyselect_1.1.1 xfun_0.22 bslib_0.2.4 haven_2.4.1
## [5] lattice_0.20-41 colorspace_2.0-0 vctrs_0.3.7 generics_0.1.0
## [9] htmltools_0.5.1.1 yaml_2.2.1 utf8_1.2.1 rlang_0.4.11
## [13] jquerylib_0.1.4 pillar_1.6.0 withr_2.4.2 glue_1.4.2
## [17] DBI_1.1.1 dbplyr_2.1.1 readxl_1.3.1 modelr_0.1.8
## [21] lifecycle_1.0.0 cellranger_1.1.0 munsell_0.5.0 gtable_0.3.0
## [25] rvest_1.0.0 evaluate_0.14 labeling_0.4.2 knitr_1.33
## [29] fansi_0.4.2 highr_0.9 Rcpp_1.0.6 broom_0.7.6
## [33] backports_1.2.1 scales_1.1.1 jsonlite_1.7.2 farver_2.1.0
## [37] fs_1.5.0 hms_1.0.0 digest_0.6.27 stringi_1.5.3
## [41] grid_4.0.2 mathjaxr_1.4-0 cli_2.5.0 tools_4.0.2
## [45] magrittr_2.0.1 sass_0.3.1 crayon_1.4.1 pkgconfig_2.0.3
## [49] ellipsis_0.3.1 xml2_1.3.2 reprex_2.0.0 lubridate_1.7.10
## [53] assertthat_0.2.1 rmarkdown_2.7 httr_1.4.2 rstudioapi_0.13
## [57] R6_2.5.0 nlme_3.1-148 compiler_4.0.2