ggscatter displays a positive pearson correlation coefficient in kaggle notebook instead of a negati-CodePudding

I'm currently working on my first kaggle notebook and faced the issue stated in the title. The minus is only missing in kaggle, as you can see in the pictures. As I want to demonstrate a negative correlation, it is important to see the R is - 0.6.

plot in kaggle notebook
plot in RStudio

Here is the code I use in both pictures:

ggscatter(activity_sleep, x = "TotalMinutesAsleep", 
          y = "SedentaryMinutes", shape = 21, add = "loess",
          add.params = list(color = "blue", fill = "darkgrey"),
          conf.int = TRUE, cor.coef = TRUE, cor.method = "pearson")   
          labs(title="Sedentary Minutes vs. Minutes Asleep")

Is there a way to fix this?

CodePudding user response：

This looks like a kaggle bug in the graphics driver. Running this code in kaggle also leaves out the minus sign:

plot(1, main = expression(-1))

On the other hand, this works:

plot(1, main = "-1")

According to the docs for ggpubr::ggscatter(), you should be able to choose to display the stats in text rather than using an expression, but it didn't work when I tried this:

activity_sleep <- tibble::tibble(TotalMinutesAsleep = rnorm(20),
   SedentaryMinutes = rnorm(20) - TotalMinutesAsleep)
ggpubr::ggscatter(activity_sleep, x = "TotalMinutesAsleep", 
                  y = "SedentaryMinutes", shape = 21, add = "loess" ,
                  add.params = list(color = "blue", fill = "darkgrey"), 
                  conf.int = TRUE, 
                  cor.coef = TRUE, cor.method = "pearson", 
                  cor.coef.args = list(output.type = "text"))  
    labs(title="Sedentary Minutes vs. Minutes Asleep")

Bugs everywhere!

CodePudding user response：

@user2554330 Thank you for your help! Based on your input I figured out how to solve the issue following this doc.

The code now looks like this:

ggscatter(activity_sleep, y = "Calories", x = "SedentaryMinutes", 
      shape = 20, 
      add = "reg.line", 
      add.params = list(color = "blue", fill = "darkgrey"), 
      conf.int = TRUE)   
      stat_cor(
          aes(label = ..r.label..), label.x = 3,
          method = "pearson",
          output.type = "text")

If output.type = "expression", kaggle does not show the minus. All other options work.