# Pearson Correlation Coefficient

• Sep 10th 2008, 02:09 AM
yiklim
Pearson Correlation Coefficient
I am involved in a project investigating if knowledge on a food product is associated to its consumption in a certain population.

I have classified the knowledge into low, moderate and high based on the subjects' scores obtained from questionnaires. To know if they are associated, I used a chi-squared test.

But I was suggested to use a Pearson Correlation Test.

So here are my questions:

1. Can this test be used on variables that are not quantifiable? If no, could I assign an arbitrary value to my variables so that I can use this test? For example, using 1 for low knowledge, 2 for moderate knowledge and 3 for high knowledge.

2. I read that the pearson correlation coefficient measures the degree and direction of a linear relationship between two variables. Do I need to confirm if the relationship is linear in the first place before calculating the pearson correlation coefficient? If yes, how? Can I use this coefficient to indicate a relationship (similar to a chi-squared test), not necessarily a linear one, between two variables?

3. Is there such a thing as pearson correlation coefficient test, which will generate a p-value like other significance tests? Is yes, why would one be interested in the p-value, which I assume only indicates if there's any linear relationship/relationship as opposed to the coefficient itself, which not only indicates the relationship, but also its strength?