Regression is derived from correlation: once the correlation is known, the regression equation can be obtained. Theregression equation is used for predicting the value of one variablefrom a specified value of another, correlated variable.

There are several differences between correlations and regressions. In acorrelation there is no reference to causation. Neither variable is consideredto be the independent or the dependent variable. They are two separatevariables that co-relate, or have a specific relatinship. In a regression onevariable is considered to be the predictor (independent) variable, and theother is the criterion (dependent) variable. Thus, when we have theresults of the regression analysis, we can say that one variable causes, or isresponsible for a certain percentage of variation in the other variable.

Another reason to use a regression analysis rather than a correlationformula (assuming your data meet the assumptions required for regression) isthat the results of a regression are much more generalizable than those from acorrelation.

For example, we might collect data related to the number of hours studentsstudy (the predictor variable) and their grades in a course (the criterionvariable). If we perform a regression analysis and find that there is asignificant relationship we would be able to say that hours of study contributeto the student's performance and we could predict their grades if we knew thehours they studied.

We can also perform a multiple regression. In this instance we will haveseveral predictor variables and one criterion variable and we are trying todiscover how much each predictor variable contributes to the variation in thecriterion measure. For example we might collect data about the number ofhours a student studies, the learning strategy they use, their age, and theirIQ (all predictor variables), and their course grades (the criterion variable). By performing a multiple regression analysis we could discover which predictorvariable was most responsible for the variation in grades. In this way wewould know that if learning strategy the student used accounted for the mostvariability in grades, we should use that variable as our most importantpredictor.