Plotting pre-post results in R

A quick code to create a useful plot to visualise pre- and post-test results. I was inspired by this post, this post, and this post. A scatterplot between pre-test and post-test scores for the control and the training group is presented. The straight black lines represent correlations between pre-test and post-test scores within the two groups. The straight lightgrey line represents the same values at pre-test and post-test –> Subjects above the line display a better performance at post-test compared to pre-test. At the margins, the boxplots are reported as well as the mean (i.e., diamond) and 95% confidence intervals (i.e., error bars). In the case of the post-test scores (plot on the right margin), the adjusted means (i.e., black triangles) and associated 95% confidence intervals (i.e., error bars) are reported. The table reports the results for the ANCOVA and its assumptions. In this example, participants in the training group show an improvement in a Working Memory task.
# Seed
set.seed(42)

# Packages
library(ggplot2)
library(Hmisc)
library(cowplot)
library(mvtnorm)
library(reshape2)
library(ggpubr)
library(ggExtra)
library(gridExtra)
library(car)
library(apaTables)
library(BayesFactor)
library(dplyr)
library(effects)


# Generate some correlated values
Scores <- c(rmvnorm(n=100,mean=c(100,100),sigma=matrix(c(15,0.8*sqrt(100), 
0.8*sqrt(100),15),2,2))) 

# Add improvement for the training group
improvement <- c(rep(0,100), rep(10, 50), rep(0,50))
Scores <- Scores + improvement

# Groups
Groups <- rep(rep(c("Training", "Control"), each=50),2)

# Session
Session <- rep(c("Pretest", "Posttest"), each = 100)

# Subject ID
Subject <- as.factor(rep(1:100,2))

# Long format
df_long <- data.frame(Subject, Session, Groups, Scores)

# Wide format
df_wide <-  dcast(df_long, Subject + Groups ~ Session, value.var="Scores")

# ANCOVA

# set orthogonal contrasts
contrasts(df_wide$Groups)=contr.poly(length(unique(df_wide$Groups))) 

# Assumption 1: equality of slopes–interaction is not significiant.
res_int <- aov(Posttest ~ Pretest + Groups + Groups:Pretest, data = df_wide)
pvalue_int <- summary(res_int)[[1]][["Pr(>F)"]][3]
res_int_bf <- lmBF(Posttest ~ Pretest + Groups + Groups:Pretest, data = df_wide, whichRandom = "Subject")
res_maineff_bf <- lmBF(Posttest ~ Pretest + Groups, data = df_wide, whichRandom = "Subject")
bf_int <- extractBF(res_int_bf/res_maineff_bf, onlybf = T)

# Assumption 2: linearity of slopes
gp <- group_by(df_wide, Groups)
cor_pvalues <- as.data.frame(dplyr::summarize(gp, cor.test(Pretest, Posttest)$p.value))
names(cor_pvalues)[2] <- "pval"
cor_bf <- as.data.frame(dplyr::summarize(gp, extractBF(correlationBF(y = Pretest, x = Posttest), onlybf = T)))
names(cor_bf)[2] <- "bf"

# Assumption 3: Equality of the groups on the covariate
res_comp_cov <- aov(Pretest ~ Groups, data = df_wide)
pvalue_comp_cov <- summary(res_comp_cov)[[1]][["Pr(>F)"]][1]
res_comp_cov_bf <- anovaBF(Pretest ~ Groups, data = df_wide, whichRandom = "Subject")
bf_comp_cov <- extractBF(res_comp_cov_bf, onlybf = T)

# Assumption 4: Homogeneity of variance
res_omvar <- leveneTest(Posttest ~ Groups, center = median, data = df_wide)
pvalue_omvar <- res_omvar$`Pr(>F)`[1]

# Results ANCOVA
res_anc <- aov(Posttest ~ Pretest + Groups, data = df_wide)
res_anc_an <- Anova(res_anc, type=3)
pvalue_anc <- res_anc_an$`Pr(>F)`[3]
res_anc_bf <- lmBF(Posttest ~ Pretest + Groups, data = df_wide, whichRandom = "Subject")
res_effcov_bf <- lmBF(Posttest ~ Pretest, data = df_wide, whichRandom = "Subject")
bf_anc <- extractBF(res_anc_bf/res_effcov_bf, onlybf = T)
pvalue_shap_anc <- shapiro.test(resid(res_anc))$p.value

# Adjusted means
adj_means <- as.data.frame(effect("Groups", res_anc))

# Table with results
result_text <- c("Interaction Group x Pretest",
                 paste("Linearity: Pretest / Posttest for", cor_pvalues$Groups[1]),
                 paste("Linearity: Pretest / Posttest for", cor_pvalues$Groups[2]),
                 "Pretest ~ Group",
                 "Homogeinity variance",
                 "ANCOVA: effect of group",
                 "Normality Residuals ANCOVA")

p_values <- c(pvalue_int,
              cor_pvalues$pval[1],
              cor_pvalues$pval[2],
              pvalue_comp_cov,
              pvalue_omvar,
              pvalue_anc,
              pvalue_shap_anc)

bfs_text <- c("BF: interaction/main effects",
              "BF10",
              "BF10",
              "BF10",
              NA,
              "BF: Pretest+Group/Pretest",
              NA)

bfs <- c(bf_int,
         cor_bf$bf[1],
         cor_bf$bf[2],
         bf_comp_cov,
         NA,
         bf_anc,
         NA)

table_res <- data.frame(result_text, format.pval(p_values, digits = 3), bfs_text, formatC(bfs, digits = 3))
row.names(table_res)<-NULL
names(table_res) <- c("Effect", "p-value", "BF comparison", "BF")
tab_res <- ggtexttable(table_res, rows=NULL)

# Main scatterplot
pmain <- ggplot(data = df_wide, aes(x=Pretest, y=Posttest, shape=Groups, fill=Groups)) +
  geom_abline(slope = 1, col="grey") +
  geom_point(size = 3) +
  scale_shape_manual(values = c(21,22)) +
  scale_fill_manual(values = c("white", "grey70")) +
  stat_smooth(method = "lm", se = F, col="Black") +
  theme_classic(base_size = 12) +
  theme(axis.title.x = element_text(face="bold"),
        axis.title.y = element_text(face="bold"),
        legend.title = element_text(face="bold"),
        strip.text = element_text(face="bold"),
        axis.text.x = element_text(colour = "Black"),
        axis.text.y = element_text(colour = "Black"),
        axis.ticks.x = element_line(colour = "Black", size = 0.2),
        axis.ticks.y = element_line(colour = "Black", size = 0.2),
        axis.line.x = element_line(size = 0.2),
        axis.line.y = element_line(size = 0.2),
        strip.background = element_rect(size = 0.2),
        panel.border = element_rect(size = 0.2,color="black", fill=NA),
        plot.title = element_text(face="bold", hjust = 0.5),
        legend.position="right",
        panel.grid.major = element_line(size = (0.1), colour="grey"),
        panel.grid.minor = element_blank()) +
  ggtitle("Working Memory")

# Boxplot for pre-test
xbox <- axis_canvas(pmain, axis = "x", coord_flip = TRUE) + 
  geom_boxplot(data = df_wide, aes(y = Pretest, x = Groups, fill=Groups), outlier.size = 1) + 
  stat_summary(data = df_wide, aes(y = Pretest, x = Groups), fun.data = "mean_cl_normal",colour="black",geom="errorbar", size=0.5, width=0.2) +
  stat_summary(data = df_wide, aes(y = Pretest, x = Groups), fun.y ="mean", geom="point", shape=5, size=3) + 
  scale_x_discrete() +
  coord_flip() +
  scale_fill_manual(values = c("white", "grey70"))

# Boxplot for post-test
pd=0.2
ybox <- axis_canvas(pmain, axis = "y") + 
  geom_boxplot(data = df_wide, aes(y = Posttest, x = Groups, fill=Groups), outlier.size = 1) +
  stat_summary(data = df_wide, aes(y = Posttest, x = Groups), 
               fun.data = "mean_cl_normal",colour="black",geom="errorbar", size=0.5, width=0.2, positio=position_nudge(x = -pd)) +
  stat_summary(data = df_wide, aes(y = Posttest, x = Groups), fun.y ="mean", geom="point", shape=5, size=3, positio=position_nudge(x = -pd)) + 
  scale_x_discrete() +
  scale_fill_manual(values = c("white", "grey70")) +
  geom_point(data = adj_means, aes(y = fit, x = Groups), color = "Black", size = 3, shape=17, position = position_nudge(pd)) +
  geom_errorbar(data = adj_means, aes(y = fit, ymin = lower, ymax=upper, x = Groups), size=0.5, width=0.2, position = position_nudge(pd))

# Combine plots
p1 <- insert_xaxis_grob(pmain, xbox, grid::unit(1, "in"), position = "top")
p2 <- insert_yaxis_grob(p1, ybox, grid::unit(1, "in"), position = "right")
# Show the plot
ggarrange(p2, tab_res, nrow=2, heights = c(3,1))

Comments

Post a Comment

Popular posts from this blog

JASP - Statistical software

  JASP , which stands for "Jeffreys's Amazing Statistics Program", is a free and open-source software designed for statistical analysis. Developed by a team of statisticians and programmers, JASP provides an easy-to-use graphical interface for conducting a variety of statistical tests and analyses. The software features a wide range of statistical models, including Bayesian and frequentist approaches, making it a versatile tool for data analysis in fields such as psychology, social sciences, and education.  The software can be downloaded here:  https://jasp-stats.org/download/    Tutorials here:  https://jasp-stats.org/how-to-use-jasp/   Youtube videos here:  https://www.youtube.com/watch?v=bVR3LrmfNYA&list=PLVI_iGT5ZuRnIqpav1UaadEJmj2e8QHzn  
Read more

Cold shower

Image
Cold showers have been used for centuries as a way to invigorate the body and mind. While the idea of taking a cold shower may seem daunting to some, the benefits of cold showers are numerous and can lead to improved physical and mental health. In this blog post, we'll explore some of the benefits of cold showers and why you should consider incorporating them into your daily routine. Boosted Immune System Cold showers can help boost your immune system by increasing the production of white blood cells. These cells are responsible for fighting off infections and diseases, making them essential for maintaining good health. Cold showers also stimulate the lymphatic system, which is responsible for removing waste and toxins from the body. Improved Circulation Cold water causes your blood vessels to constrict, which helps to improve circulation. This can be particularly beneficial for those who suffer from poor circulation or conditions ...
Read more

Get some morning light!

Image
The early morning hours, just after sunrise, can be a magical time of day. The light is soft, the air is cool, and the world is still relatively quiet. But did you know that there are also some significant health benefits to exposing yourself to early morning light? In this blog post, we'll explore some of the benefits of early morning light exposure and why it's worth making the effort to rise early and enjoy the morning light. Improved sleep patterns.  One of the most important benefits of early morning light exposure is that it can help to regulate your sleep patterns. Exposure to natural light in the morning helps to reset your body's internal clock, making it easier to fall asleep at night and wake up in the morning. This is because exposure to morning light helps to suppress the production of melatonin, the hormone that regulates sleep, and increase cortisol, a hormone that helps you feel more awake and alert. Boosted mood.  Exposure to early morning light has also be...
Read more