Add Spearman Correlation Coefficient and RMSE in a plot, and change the plot size

Question

I've got the following dataset:

structure(list(Count = 1:14, GW = c(0.08, 0.04, 0.35, 0.54, 0.39, 
0.94, 0.51, 0.01, 0.44, 0.63, 0.14, 0.79, 0.43, 0.73), Pz1 = c(-2.459826614, 
-2.905007821, -2.241113224, -1.549264338, -1.761962438, -1.282612221, 
-0.428828702, 1.659042479, 2.63518648, 3.076022461, 1.886216859, 
0.124473561, -1.025720789, -1.969461882), Pz2 = c(-2.916090168, 
-3.262459435, -2.455396094, -1.488106654, -0.417171756, -1.781014095, 
-0.605012986, 1.037062685, 1.977265974, 2.587846362, 2.499228916, 
1.0852274, -0.503736287, -1.829562138), Pz3 = c(-2.507944967, 
-3.718722989, -2.812847708, -1.702389524, -0.356014073, -0.436223413, 
-1.10341486, 0.860878402, 1.355286181, 1.929925855, 2.011052817, 
1.698239458, 0.457017552, -1.307577636), Pz4 = c(-2.526729696, 
-3.310577787, -3.269111262, -2.059841138, -0.570296943, -0.375065729, 
0.241375822, 0.362476528, 1.179101897, 1.307946062, 1.35313231, 
1.210063358, 1.07002961, -0.346823797), Pz5 = c(-3.284551238, 
-3.329362517, -2.86096606, -2.516104692, -0.927748557, -0.5893486, 
0.302533506, 1.70726721, 0.680700023, 1.131761778, 0.731152517, 
0.552142851, 0.58185351, 0.266188261), Pz6 = c(-4.011896321, 
-4.087184059, -2.87975079, -2.107959491, -1.38401211, -0.946800213, 
0.088250636, 1.768424893, 2.025490705, 0.633359905, 0.554968233, 
-0.069836942, -0.076066996, -0.221987839), Pz7 = c(-4.769878994, 
-4.814529142, -3.637572331, -2.12674422, -0.975866909, -1.403063767, 
-0.269200978, 1.554142023, 2.086648388, 1.978150587, 0.05656636, 
-0.246021225, -0.698046789, -0.879908346)), class = "data.frame", row.names = c(NA, 
-14L))

And I am running this code to automatically save the plots in a folder:

library(tidyverse)
library(ggplot2)
#~~~ step 1: transform the data from 'wide' format to 'long' format ~~~

my_data %>%
    pivot_longer(names_to = 'key', values_to = 'value',-c(Count, GW)) %>%
    {. ->> my_data_long}


#~~~ step 2: write a function to plot the data and save the plot ~~~

saveplot_function <- function(i){
   
     my_plot <- my_data_long %>%  

      filter(  
        key == i
      ) %>%

     ggplot(aes(x=Count)) + 
       geom_line(aes(y = scale(GW)), color = "blue") + 
       geom_line(aes(y = scale(value)), color="red") +
          ylab(i)     
          ggsave(paste0('my_location/my_folder/', i, '.png'))
  }


#~~~ step 3: loop through all the values of 'key' (column names in the wide format data) ~~~

 for(i in unique(my_data_long$key)){
    
    saveplot_function(i)
    
  }

This code does the job, but I'd need to add a box (or just the text) showing the Spearman Correlation Coefficient and the RMSE for each plot (or better say for the two time series displayed within each plot). Also, the plots are saved in a squared shape, and I would like them to be rectangular instead.

Something like this would be great:

Could you kindly help me to rewrite the code accordingly? I am not an R expert at all and this code has been modified from an old code that someone kindly wrote for me ( and it took me ages to make it work for the purpose I am using it for), so I would very much appreciate it if you could rewrite it (sorry.).

Additionally, as you can see from the code above, I used scale() . What I am trying to do is to standardise the plotted data using Z-Scores (ie, Z = (x - mean(x)) / Standard_Deviation(x) ). Can you kindly confirm (or reject) that scale() standardises the time series using Z-Scores?

Answer 1

I am a little unclear about some parts of your question, but hopefully this will help. According to the documentation, scale() does the z-score transformation with numeric data because the center value is the mean and the scale value is the standard deviation.

Adding text to the plot can be done with annotate() and providing text and coordinates. I used two separate annotations, one for correlation that also used bquote() to get the Greek letter rho and the other for RMSE. I used the max values of GW, Value, and Count to anchor the text. This might need to be refined. I also computed the RMSE and correlation of the unscaled series.

A sample plot are provided below. I would check to make sure that things are being calculated the way you expect.

my_data = structure(list(Count = 1:14, GW = c(0.08, 0.04, 0.35, 0.54, 0.39, 
                                    0.94, 0.51, 0.01, 0.44, 0.63, 0.14, 0.79, 0.43, 0.73),
               Pz1 = c(-2.459826614, -2.905007821, -2.241113224, -1.549264338, -1.761962438, -1.282612221, 
                       -0.428828702, 1.659042479, 2.63518648, 3.076022461, 1.886216859,0.124473561, -1.025720789, -1.969461882),
               Pz2 = c(-2.916090168,-3.262459435, -2.455396094, -1.488106654, -0.417171756, -1.781014095, 
                       -0.605012986, 1.037062685, 1.977265974, 2.587846362, 2.499228916,1.0852274, -0.503736287, -1.829562138),
               Pz3 = c(-2.507944967,-3.718722989, -2.812847708, -1.702389524, -0.356014073, -0.436223413, 
                       -1.10341486, 0.860878402, 1.355286181, 1.929925855, 2.011052817,1.698239458, 0.457017552, -1.307577636),
               Pz4 = c(-2.526729696,-3.310577787, -3.269111262, -2.059841138, -0.570296943, -0.375065729, 
                       0.241375822, 0.362476528, 1.179101897, 1.307946062, 1.35313231,1.210063358, 1.07002961, -0.346823797),
               Pz5 = c(-3.284551238,-3.329362517, -2.86096606, -2.516104692, -0.927748557, -0.5893486, 
                       0.302533506, 1.70726721, 0.680700023, 1.131761778, 0.731152517,0.552142851, 0.58185351, 0.266188261),
               Pz6 = c(-4.011896321,-4.087184059, -2.87975079, -2.107959491, -1.38401211, -0.946800213, 
                       0.088250636, 1.768424893, 2.025490705, 0.633359905, 0.554968233,-0.069836942, -0.076066996, -0.221987839),
               Pz7 = c(-4.769878994,-4.814529142, -3.637572331, -2.12674422, -0.975866909, -1.403063767, 
                       -0.269200978, 1.554142023, 2.086648388, 1.978150587, 0.05656636,-0.246021225, -0.698046789, -0.879908346)),
          class = "data.frame",
          row.names = c(NA,-14L))


library(tidyverse)
library(ggplot2)
#~~~ step 1: transform the data from 'wide' format to 'long' format ~~~

my_data %>%
  pivot_longer(names_to = 'key', values_to = 'value',-c(Count, GW)) %>%
  {. ->> my_data_long}


#~~~ step 2: write a function to plot the data and save the plot ~~~

saveplot_function <- function(i){
  
  # get data for key i
  df_temp <- my_data_long %>%  
    filter(
      key == i
    )
  
  # calculate RMSE and correlation
  # adjust calculation as needed
  rmse <- round(mean((df_temp$GW-df_temp$value)^2),3)
  correl <- round(cor(df_temp$GW,df_temp$value),3)
  
  my_plot <- my_data_long %>%  
    
    filter(
      key == i
    ) %>%
    
    ggplot(aes(x=Count)) + 
    geom_line(aes(y = scale(GW)), color = "blue") + 
    geom_line(aes(y = scale(value)), color="red") +
    ylab(i)+
    annotate(geom="text", x=max(df_temp$Count)-2,
             y=max(df_temp$value,df_temp$GW),
             label=bquote(paste(," rho (",rho,") =",.(correl),sep="")),#expression(paste("rho (", rho, ") = ",as.character(correl),sep='')),
             color="black")+
    annotate(geom="text", x=max(df_temp$Count)-2,
           y=max(df_temp$value,df_temp$GW)-0.5,
           label=paste('RMSE = ',rmse,sep=''),
           color="black")
  
  ggsave(paste0("my_location/my_folder/", i, '.png'),
         width=6,
         height=4,
         units='in')
}


#~~~ step 3: loop through all the values of 'key' (column names in the wide format data) ~~~

for(i in unique(my_data_long$key)){
  
  saveplot_function(i)
  
}