Correlation coefficient
Hello to everibody,
I have noted that, sometimes, the correlation coefficients given by JASP differ from those calculated with other softwares. It is not just an issue of rounding because the difference affects not only the last digit but the last two.
Does JASP apply some statistical correction vector that I do not know?
As example, I created the following two vectors and calculated their correlation coefficients with JASP and MATLAB:
X = 1, 2, 3, 4, 5, 6, 7, 8, 9 10
Y = 3, 13, 9, 9, 11, 11, 12, 16, 17, 18
Please, note that I deliberately chosen integer data for avoinding issues about the number of decimal digits retained by the two softaware.
When all 10 pairs of data are used both softewares gave the same value for Pearson's correlation coefficent, 0.852. When, however, only the LAST 5 data pairs (X ranging from 6 to 10) were used the two coefficient differed. MATLAB gave 0.965 while JASP gave 0.956.
OK, the difference is small, compared to the confidence interval, but I do not know why the two expectation values differ. May someone explain it?
Best regards.
Leonardo_C
Comments
Maybe you somehow messed up the data in the JASP file? With JASP, I get 0.965. See below.
R
Hi Andersony3k,
you are right. I made an error in applying the filter. I used larger or equal 5 instead of larger of 5.
I will check the other cases where such thing happened. Maybe there was an error in data inport (though, theoretically, the source file should have been the same).
Thanks very much for your help.
Leonardo-C