My friend has a dataset of two variables of her research. Both of scores of the variables move in the same direction, but no correlation between them.
How can this is happened?
Can anyone give me explanation of this issue and reference i could read to explain this issue?
Thank you in advance.
Best regards,
Riangga.
They can't. Unless you made a hand calculation error or misread the printout from a data analytic program, the only possibility is that the two variables "seem" to be moving in tandem on casual inspection because that's what your colleague expects. The phenomenon is known as illusory correlation and was first described by Loren and Jean Chapman 50 years ago. Here is the reference for one of their classic papers: Chapman, L. J., & Chapman, J. P. (1969). Illusory correlation as an obstacle to the use of valid psychodiagnostic signs. Journal of Abnormal Psychology, 74(3), 271-280.
They can't. Unless you made a hand calculation error or misread the printout from a data analytic program, the only possibility is that the two variables "seem" to be moving in tandem on casual inspection because that's what your colleague expects. The phenomenon is known as illusory correlation and was first described by Loren and Jean Chapman 50 years ago. Here is the reference for one of their classic papers: Chapman, L. J., & Chapman, J. P. (1969). Illusory correlation as an obstacle to the use of valid psychodiagnostic signs. Journal of Abnormal Psychology, 74(3), 271-280.
John Hyland
Yes, it was.
r value was .029 with insignificant p-value .438(p>.05)
That r value is very low. Such a value would indicate a high likelihood that no relationship exists.
Arrange your data of any one variable in ascending order, along with change in order of the values of the other variable corresponding to their values in first sorted variable. Now what pattern do you see ?
You can also create ranks of the data to see if their is any large discrepancy in rank orders. This can be done if data is small in number. But for larger dataset, divide it to see the pattern again. For example, if you have sample size of 300 then take 3 subsets of 50 participants, and then do the rank ordering of each subset to understand this data.
This above suggestion applies only if you want to verify your doubt that both variables seem to covary. You need to recheck your assumption and I have mentioned two common ways to do that.
There is a possibility that there is some confounding variable which is lowering the correlation between the two variables which seemingly are moving in the same direction.
Surely the linear model relating the two variables is not adequate and the residuals are very large. Try a monotonic regression model.
- Badges
- Science topic