In tbates/meta.now: Up-to-date meta-analysis (initially of GxSES effects on cognitive ability).

Meta-analysis of GxSES effects on Cognitive Ability

library("meta.now");
library("RefManageR")
library("metaSEM")
library("ggplot2")
library("ggthemes")
library("cowplot"); # ggplot enhancer

me = "GxSES"
genbib = RefManageR::ReadBib(system.file("extdata", "general.bib", package = "meta.now"))

# Assumes mn_update_data() has been run
# Read in the overall file
df = gdata::read.xls(paste0(path, "/", "data_GxSES.xlsx"), stringsAsFactors = FALSE)
str(df)

# 2. open the bib for this project
bib = ReadBib(system.file(paste0("extdata/", me), paste0("bib_", me, ".bib"), package = "meta.now"), check = FALSE)
BibOptions(check.entries = FALSE, style = "markdown", cite.style = "authoryear", bib.style = "numeric")

Introduction

r AutoCite(bib, "Bates2013", before = "e.g., ")

Results

hist(df$a1)
mean(df$a1)
weighted.mean(df$a1  , 1/(df$k   * df$a1_SE^2)  , na.rm = TRUE)
weighted.mean(dfUS$a1, 1/(dfUS$k * dfUS$a1_SE^2), na.rm = TRUE))
weighted.mean(dfEU$a1, 1/(dfEU$k * dfEU$a1_SE^2), na.rm = TRUE))

# Two level unconditional (Random Effects meta analysis)
rmeta = meta(y = a1,v = a1_SE^2, data = df)
summary(rmeta)


# Three level unconditional (intercept only)
Model0 = meta3(y = a1, v = a1_SE^2, cluster = DataID, data = df, model.name = "3 level model")
summary(Model0)

# Three level US vs. Other
Model1 = meta3(y = a1, v = a1_SE^2, x = US, cluster = DataID, data = df, model.name = "3 level model")
summary(Model1)

# Weighted histogram
p = ggplot(df, aes(x = a1, weight = 1/(k * a1_SE^2))) + geom_histogram(col="black", fill = "light grey", bins = 15)
p = p + labs(x = "Effect size", y = "Weighted count") + xlim(c(-.2, .2))
p = p + theme_base() + theme(plot.background = element_rect( color = "white"))
# p + theme_bw()

# Weighted histogram US and not US (weight = reciprocal sampling variance)
plotUS = ggplot(dfUS, aes(x = a1, weight = 1/(a1_SE^2))) + geom_histogram(col = "black", fill = "light grey", binwidth = .025)
# plotUS = plotUS + ggtitle("United States")
plotUS = plotUS + labs(x = "Effect size", y = "Weighted count")
plotUS = plotUS + xlim(c(-.3, .3))
# plotUS = plotUS + theme_bw()
plotUS = plotUS + theme_base() + theme(plot.background = element_rect( color= "white"))
plotUS = plotUS + theme(plot.title = element_text(hjust = 0.5))

plotEU = ggplot(dfEU, aes(x = a1, weight = 1/(a1_SE^2))) + geom_histogram(col = "black",fill = "light grey", binwidth = .025)
# plotEU = plotEU + ggtitle("Europe and Australia")
plotEU = plotEU + labs(x = "Effect size", y= "Weighted count")
plotEU = plotEU + xlim(c(-.3, .3))
# plotEU = plotEU + theme_bw()
plotEU = plotEU + theme_base() + theme(plot.background = element_rect(color = "white"))
plotEU = plotEU + theme(plot.title = element_text(hjust = 0.5))

# both
p = ggplot(df, aes(x = a1, weight = 1/(a1_SE^2), fill = factor(US))) + geom_histogram(alpha = 0.2, position = "identity", bins=15)
p = p + labs(x = "Effect size", y= "Weighted count")
p = p + xlim(c(-.3, .3))
p = p + theme_bw()
p = p + theme_base() + theme(plot.background = element_rect(color = "white"))
p = p + theme(plot.title = element_text(hjust = 0.5))

cowplot::plot_grid(plotUS, plotEU, labels = c("United States", "Europe and Australia"), align = "v")

metafor::forest(rma(yi = a1, sei = a1_SE, data=df[df$US==1,]))
metafor::forest(rma(yi = a1, sei = a1_SE, data=df[df$US==0,]))


, annotate=TRUE, showweights=FALSE,
       xlim, alim, clim, ylim, at, steps=5,
       level=95, refline=0, digits=2L, width,
       xlab, slab, ilab, ilab.xpos, ilab.pos,
       subset, transf, atransf, targs, rows,
       efac=1, pch=15, psize, col, lty,
       cex, cex.lab, cex.axis, annosym, ...)

# Weighted histogram US and not US (weights = reciprocal sampling variance and 1/k)

plotUS = ggplot(dfUS, aes(x = a1, weight = 1/(k*a1_SE^2)))
plotUS = plotUS + geom_histogram(col = "black",fill = "light grey")
plotUS = plotUS + labs(x = "Effect size", y = "Weighted count")
plotUS = plotUS + xlim(c(-.3, .3))
plotUS = plotUS + theme_base()
plotUS = plotUS + theme(plot.background = element_rect( color= "white"))
plotUS = plotUS + ggtitle("United States")
plotUS = plotUS + theme(plot.title = element_text(hjust = 0.5))

plotEU = ggplot(dfEU,aes(x=a1,weight=1/(k*a1_SE^2)))
plotEU = plotEU + geom_histogram(col="black",fill="light grey")
plotEU = plotEU + labs(x="Effect size", y="Weighted count")
plotEU = plotEU + xlim(c(-.3,.3))
plotEU = plotEU + theme_base()
plotEU = plotEU + theme(plot.background = element_rect( color= "white"))
plotEU = plotEU + ggtitle("Europe and Australia")
plotEU = plotEU + theme(plot.title = element_text(hjust = 0.5))

multiplot(plotUS,plotEU)

Discussion

Data References

PrintBibliography(bib, .opts = list(check.entries = FALSE, sorting = "ynt"))

References for packages and methods used

PrintBibliography(genbib)

Citing

1. In text r Citet(bib, "loh")
2. In parentheses r AutoCite(bib, "baez/online", before = "e.g., ").
3. Without citing = r NoCite(bib = bib, title = "CTAN")
4. methods
   1. by key
   2. using SearchBib() 1 by type r Citet(bib, bibtype = "Report", .opts = list(hyperlink = "to.doc")).
      1. From another bibliography: r AutoCite(bib2, title = "binary longitudinal data").
      2. By author: r TextCite(bib, author = "Aristotle").
      3. By other fields: r TextCite(bib, eprinttype = "arxiv").

Setting options

• Setup chunk at the start
• Anywhere with BibOptions
• List in .opts argument to cite functions.
• Can mix "authoyear" and "numeric"style.

hyperlinking

1. default = open document in a new window (see ?open.BibEntry).
2. no link r AutoCite(bib, location = "Uppsala", .opts = list(hyperlink = FALSE)).

Plots

plot(cars)

plot(cars)

tbates/meta.now documentation built on Sept. 13, 2019, 6:14 a.m.