1

Since Pearson correlation coefficient deal with linear correlation between two variables, I wonder if it is correct to show the variables on a log scale axis. For example, in the figure below, the R value is high and P value is low. So assuming alpha=0.05, we can say that there is a strong positive correlation between Y and X. However, the Y is shown on a log scale axis. Does that matter?

enter image description here

UPDATE:

If I don't use the log scale on the y-axis, the results is like below. As you can see, the R and P values are different.

enter image description here

mahmood
  • 193
  • 7
  • 5
    The way the data is shown graphically doesn't really matter, the input you used to calculate the pearson coeff does. If the correlation is calculated on log(Y) and not Y, then you can't say that Y and X have a linear relationship – Mayeul sgc May 26 '21 at 09:58
  • Please see the updated post. It seems that there (not) using log scale alters the chart. – mahmood May 26 '21 at 12:35
  • if your results changed then you modified your input in the pearson coeff calculation. It impacts the coefficient indeed, please see [this](https://stats.stackexchange.com/questions/127121/do-logs-modify-the-correlation-between-two-variables) answer for more info – Mayeul sgc May 27 '21 at 00:18

0 Answers0