what is a confounding variable?

[prunemuffins]

can someone please tell me what it is? can you also put down a reference please?

[Irgumburgum]

I do not know.

[Victor]

A confounding variable is one which creates a spurious relationship between two or more other variables.
So variable A may correlate closely with variable B; and variable B correlates closely with variable C.
So a data set between A and C will show correlation; even if there is no rational, physical or theoretical connection between them.
Then B is the confounding variable.

A classic case is the apparently high correlation between strawberry sales and the incidence of countryside accidents. It could seem that eating strawberries causes people to have accidents. Or perhaps the occurrence of accidents in the countryside leads people to buy more strawberries as a consolation food.

The true explanation is that higher temperatures and pleasant weather in the summer bring about the strawberry season in markets and stores. Also, these climatic conditions encourage people to take vacations in the country, or just family walks, or serious orienteering.

The confounding variable is the average daily temperature.

This phenomenon is important in medical and agricultural investigations. It is always easy to see the correlation coefficient between two variables as "proof" that one thing causes another. But the worlds of epidemiology, medical statistics, and weather forecasting are overwhelmed by confounding variables. Their identification is crucial to understanding the true relationship; but very many social and medical scientists don't look very far, and assume that they've hit on a real correlation when they haven't.

The article below is a good intro to the problems and its identification.