statistics

[moonstone71]

How do i work out the statistical correlation between 2 variables?

[BadBird2]

How do i work out the statistical correlation between 2 variables?In order to answer this question appropriately, we would have to know more about your background. Your Public bio says, " ...novice writer with an interest in religion, psychology, history, politics and astrology. Mum to 5 children." I'm going to assume that you are not looking for a dissertation on mathematical techniques for curve-fitting, statistical analyses, etc., but simply want a general idea of how one can get an idea of how two variables are related. (If I am wrong, please accept my apology.)

First you must gather data to determine which variable is the independent variable and which is dependent or if they are co-dependent. If you have a conjecture, start by varying the independent variable and observing the dependent variable over as wide a range of independent variables as possible. (make a table). Take multiple readings for each independent variable. (Be careful, because their may be more than two variables, in which case you would have to do multiple runs, holding the others constant!)

If there is any question as to which variable is independent, repeat, reversing the variables in order to establish that you have chosen the right variable to be independent.

Plot the points on a graph, with independent variable along the X-axis and the dependent variable along the Y-axis. With luck, the dependent variables at each independent variable value will be identical, or close to identical (within the accuracy of your observations) and you can plot a smooth curve or line between the points by eyeball. It is OK if your smooth curve or line doesn't go exactly through the dependent variables, just so it is fairly close.

If you aren't so lucky there will be many different dependent values for each independent value creating a long skinny "cloud" of data points, in which case your smooth curve or line should go through--or near--the average value of the dependent variable at each independent variable value. Too much 'cloud' indicates that there are other variables at work that you hadn't controlled.

Finding the very best curve requires use of those statistical techniques that I mentioned above: there are many such techniques, but for rough answers they are overkill--and way too much work.

Start with a straight line: that is the most common relationship. Algebraically it can be described a Y=m*X+b {where m is the slope of the line and b is the value of Y when X=0}. There are other less common curves such as Y=m*(X squared)+b, etc.

Is that enough of an explanation for you? If not you'll have to get an appropriate book on math or statistics. If you don't let it scare you, this can be an interesting subject and it really does help us understand the world around us.