Creation of Boxplot in ggplot2

Question

I’m doing some simulations on R. By generating a data.frame and request the impression of boxplots with small sample size and number of samples boxplot is generated perfectly. The problem is when increasing the sample size and the number of samples the boxplots are not generated according to the data. To exemplify I printed my data in excel sheets and generated the boxplots from these spreadsheets, I keep receiving boxplots wrong as you can see below. I cannot find the problem!

##Box-Plot dos valores simulados dos parâmetros.

library(ggplot2)
library(dplyr)
library(readr)
df1 <- read.csv(url("https://cdn.rawgit.com/fsbmat/StackOverflow/58046aff/d1.csv"), sep = ";")
attach(df1)
df1$parametros <- factor(df1$parametros,
                         levels = c("gamma0","gamma1","gamma2","beta0","beta1","beta2","phi1", "rho"), ordered = TRUE)

g1 <- ggplot(df1, aes(x = parametros,y = val_Sim)) +
  geom_boxplot(aes(fill = parametros),alpha = .6,size = .5)+
  stat_boxplot(geom ='errorbar') +
  guides(fill=FALSE)+geom_point()+
  ggtitle("Boxplot com os valores estimados") +   xlab("Parâmetros")+
  scale_y_continuous(name = "Valores Estimados",
                     breaks = seq(-0.5, 2, 0.5),
                     limits=c(-0.5, 2))+
  theme(plot.title = element_text(hjust = 0.5))

df2 <- read.csv(url("https://cdn.rawgit.com/fsbmat/StackOverflow/58046aff/d2.csv"), sep = ";")

attach(df2)
df2$parametros <- factor(df2$parametros,
                         levels = c("gamma0","gamma1","gamma2"), ordered = TRUE)


g2 <- ggplot(df2, aes(y = val_Sim, x = parametros)) +
  geom_boxplot(aes(fill = parametros),alpha = .6,size = .5)+
  stat_boxplot(geom ='errorbar') +
  guides(fill=FALSE)+geom_point()+
  ggtitle("Boxplot com os valores estimados") +   xlab("Parâmetros") +
  scale_y_continuous(name = "Valores Estimados",
                     breaks = seq(-0.5, 2, 0.5),
                     limits=c(-0.5, 2))+
  theme(plot.title = element_text(hjust = 0.5))


df3 <- read.csv(url("https://cdn.rawgit.com/fsbmat/StackOverflow/58046aff/d3.csv"), sep = ";")
attach(df3)
df3$parametros <- factor(df3$parametros,
                         levels = c("beta0","beta1","beta2"), ordered = TRUE)

g3 <- ggplot(df3, aes(y = val_Sim, x = parametros)) +
  geom_boxplot(aes(fill = parametros),alpha = .6,size = .5)+
  stat_boxplot(geom ='errorbar') +
  guides(fill=FALSE)+geom_point()+
  ggtitle("Boxplot com os valores estimados") +   xlab("Parâmetros") +
  scale_y_continuous(name = "Valores Estimados",
                     breaks = seq(-0.5, 2, 0.5),
                     limits=c(-0.5, 2))+
  theme(plot.title = element_text(hjust = 0.5))


df4 <- read.csv(url("https://cdn.rawgit.com/fsbmat/StackOverflow/58046aff/d4.csv"), sep = ";")
attach(df4)
df4$parametros <- factor(df4$parametros,
                         levels = c("phi1", "rho"), ordered = TRUE)



g4 <- ggplot(df4, aes(y = val_Sim, x = parametros)) +
  geom_boxplot(aes(fill = parametros),alpha = .6,size = .5)+
  stat_boxplot(geom ='errorbar') +
  guides(fill=FALSE)+geom_point()+
  ggtitle("Boxplot com os valores estimados") +   xlab("Parâmetros") +
  scale_y_continuous(name = "Valores Estimados",
                     breaks = seq(-0.5, 2, 0.25),
                     limits=c(-0.5, 2))+
  theme(plot.title = element_text(hjust = 0.5))
library(gridExtra)
grid.arrange(g1,g2, g3, g4)

Answer 1

For some unknown reason, your files df2, df3 and df4 are wrong. See, for example, what happens when I calculate the median of each group, using df1, df2_original (which is yours df2) and df2_correto (I created from the df1, selecting only lines with gamma values):

df1 <- read.csv(url("https://cdn.rawgit.com/fsbmat/StackOverflow/58046aff/d1.csv"), sep = ";")
df1 %>% 
  group_by(parametros) %>% 
  summarise(mediana = median(val_Sim, na.rm = TRUE))
# A tibble: 8 × 2
  parametros   mediana
      <fctr>     <dbl>
1      beta0 0.9406746
2      beta1 0.8604181
3      beta2 0.2352544
4     gamma0 0.7085335
5     gamma1 0.8904442
6     gamma2 0.9441261
7       phi1 0.2305419
8        rho 0.2301348
df2_original <- read.csv(url("https://cdn.rawgit.com/fsbmat/StackOverflow/58046aff/d2.csv"), sep = ";")
df2_original %>% 
  group_by(parametros) %>% 
  summarise(mediana = median(val_Sim, na.rm = TRUE))
# A tibble: 3 × 2
  parametros   mediana
      <fctr>     <dbl>
1     gamma0 0.7068794
2     gamma1 0.9087855
3     gamma2 0.8771044
df2_correto <- filter(df1, parametros=="gamma0" | parametros=="gamma1" | 
  parametros=="gamma2")
df2_correto %>% 
  group_by(parametros) %>% 
  summarise(mediana = median(val_Sim, na.rm = TRUE))
# A tibble: 3 × 2
  parametros   mediana
      <fctr>     <dbl>
1     gamma0 0.7085335
2     gamma1 0.8904442
3     gamma2 0.9441261

Note that the medians of df2_correto are identical to those of df1. They are even in ascending order, as the boxplot with all parameters suggests. So I suspect there was some mistake in the creation of your original subsets, ie the files df2.csv, df3.csv and df4.csv are in some trouble.

Also, note that the function syntax filter is quite simple. It was trivial to select the lines that interested me to create the df2 only with the values of gamma0, gamma1 and gamma2. I didn’t need to create separate files for this, making the code a little more organized and easier to debug.

Therefore, by doing so I was able to obtain the desired graphics without having to appeal to . csv files other than the original file d1.csv. For this, I used the function a few more times filter package dplyr:

library(ggplot2)
library(dplyr)
library(readr)
library(gridExtra)

df1 <- read.csv(url("https://cdn.rawgit.com/fsbmat/StackOverflow/58046aff/d1.csv"), sep = ";")

df1$parametros <- factor(df1$parametros,
                         levels = c("gamma0","gamma1","gamma2","beta0","beta1","beta2","phi1", "rho"), ordered = TRUE)

g1 <- ggplot(df1, aes(x = parametros,y = val_Sim)) +
  geom_boxplot(aes(fill = parametros),alpha = .6,size = .5)+
  stat_boxplot(geom ='errorbar') +
  guides(fill=FALSE)+geom_point()+
  ggtitle("Boxplot com os valores estimados") +   xlab("Parâmetros")+
  scale_y_continuous(name = "Valores Estimados",
                     breaks = seq(-0.5, 2, 0.5),
                     limits=c(-0.5, 2))+
  theme(plot.title = element_text(hjust = 0.5))

df2_correto <- filter(df1, parametros=="gamma0" | parametros=="gamma1" | 
  parametros=="gamma2")

g2 <- ggplot(df2, aes(x = parametros,y = val_Sim)) +
  geom_boxplot(aes(fill = parametros),alpha = .6,size = .5)+
  stat_boxplot(geom ='errorbar') +
  guides(fill=FALSE)+geom_point()+
  ggtitle("Boxplot com os valores estimados") +   xlab("Parâmetros")+
  scale_y_continuous(name = "Valores Estimados",
                     breaks = seq(-0.5, 2, 0.5),
                     limits=c(-0.5, 2))+
  theme(plot.title = element_text(hjust = 0.5))

df3 <- filter(df1, parametros=="beta0" | parametros=="beta1" | parametros=="beta2")

g3 <- ggplot(df3, aes(y = val_Sim, x = parametros)) +
  geom_boxplot(aes(fill = parametros),alpha = .6,size = .5)+
  stat_boxplot(geom ='errorbar') +
  guides(fill=FALSE)+geom_point()+
  ggtitle("Boxplot com os valores estimados") +   xlab("Parâmetros") +
  scale_y_continuous(name = "Valores Estimados",
                     breaks = seq(-0.5, 2, 0.5),
                     limits=c(-0.5, 2))+
  theme(plot.title = element_text(hjust = 0.5))


df4 <- filter(df1, parametros=="phi1" | parametros=="rho")

g4 <- ggplot(df4, aes(y = val_Sim, x = parametros)) +
  geom_boxplot(aes(fill = parametros),alpha = .6,size = .5)+
  stat_boxplot(geom ='errorbar') +
  guides(fill=FALSE)+geom_point()+
  ggtitle("Boxplot com os valores estimados") +   xlab("Parâmetros") +
  scale_y_continuous(name = "Valores Estimados",
                     breaks = seq(-0.5, 2, 0.25),
                     limits=c(-0.5, 2))+
  theme(plot.title = element_text(hjust = 0.5))

grid.arrange(g1,g2, g3, g4)

Also, note that there was no need to reorder the levels of parametros for every new df, because the original date frame had already ordered them. And, finally, it is not necessary to use attach, which is a redundant function due to the syntax of ggplot and dplyr.