Isolating graphs with significant p values

Question

For starters, the data comes from the us_contagious_diseases dataset, and the packages are tidyverse and ggpubr

library(dslabs)
library(ggpubr)
library(tidyverse)
data("us_contagious_diseases")

I modified this dataset via the code below:

sdf <- us_contagious_diseases %>% filter(., disease == 'Rubella' | disease == 'Mumps') %>% transmute(., disease, count, population, state)

Then I created a boxplot comparing the numbers of Rubella and Mumps cases in each State:

sdf_plot <- ggplot(sdf, mapping = aes(x = disease, y = count)) + geom_boxplot(outlier.shape = NA) + facet_wrap('state', scales = 'free') + stat_compare_means(method = 't.test', label.y.npc = 0.8) 

The thing is, there are FIFTY ONE plots in this figure!!! That's wayyyy to huge to include in my report. More importantly, many of these comparisons don't have significant p-values. Is there a way I can pull just those plots that have a p value less than 0.01?

Answer 1

I guess you need to pre-calculate the p-values:

library(broom)
res = sdf %>% group_by(state) %>% do(tidy(t.test(count~disease,data=.)))
head(res)

# A tibble: 6 x 11
# Groups:   state [6]
  state estimate estimate1 estimate2 statistic p.value parameter conf.low
  <fct>    <dbl>     <dbl>     <dbl>     <dbl>   <dbl>     <dbl>    <dbl>
1 Alab…    125.       180.      55.4     2.46  0.0181       42.7     22.4
2 Alas…     66.1      104.      38.2     1.52  0.136        45.4    -21.7
3 Ariz…     78.6      266.     187.      0.657 0.513        68.0   -160. 
4 Arka…     84.3      113.      28.4     2.87  0.00628      45.5     25.1
5 Cali…    386.      1915.    1529.      0.540 0.592        59.3  -1046. 
6 Colo…     95.0      314.     219.      0.762 0.449        62.6   -154. 

keep = res$state[res$p.value<0.01]
[1] Arkansas             District Of Columbia Georgia             
[4] Kansas               Maryland             Nevada              
[7] Ohio 

Then plot using this filter:

sdf_plot <- ggplot(subset(sdf,state %in% keep),aes(x = disease, y = count)) + 
geom_boxplot(outlier.shape = NA) + 
facet_wrap('state', scales = 'free') + 
stat_compare_means(method = 't.test', label.y.npc = 0.8)