Cross Validated is a question and answer site for people interested in statistics, machine learning, data analysis, data mining, and data visualization. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

For example, I have three variables: A, B, C.

I analysis the partial correlation between A and B while controlling C (expressed as A~B[C]), and the partial correlation between A and C while controlling B (expressed as A~C[B]).

The result shows that the correlation coefficient of A~B[C] is 0.67, p-value < 0.001, however, the p-value of A~C[B] also < 0.001, but the correlation coefficient is small, only 0.02.

Could I explain as: at a given variable C level, A is more possibly determined by B rather than C?

share|improve this question

Keep in mind that the p-value depends on the sample size. I guess you have a large sample size thus both values are p < 0.001. My interpretation would be that A is correlated with B and less so with C. Try to compare the partial correlation results with the normal correlation between A and B/A and C.

share|improve this answer
is there any method which could be used to control/correct for the large sample sizes? I am currently running partial correlations on 5933 samples, even tiny correlations are incredibly significant. Personally I think we place too much emphasis on the p-value and we should use common-sense interpretation more often (but that is an aside, if I want to get published I have to succumb to the power of the p!) – rg255 Mar 6 '13 at 10:47
For large sizes I usually rank the tests using the test statistics since p becomes meaningless. So yes you are right, use common-sense interpretation and if your p is "too small" then you are on the safe side for publishing. – simmmons Mar 6 '13 at 14:42

At a given level of variable C, C cannot determine A because C cannot vary and thus cannot covary either.

share|improve this answer
Sorry for my misleading of phase usage. Here, 'at a give level of variable C', I meant that the variable C is between a small range (comparing the whole dataset, which variable C can vary in a very large range). What I want to know is, within a specific level of variable C, whether A is determined by B or by C. – Richard Ma Jun 8 '12 at 4:57

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.