Regression analysis is being used when there is a requirement of predicting a continuous dependent variable from a number of independent variables. Logistic regression should be used, if the dependent variable is dichotomous. If the split between the two levels of the dependent variable is close to 50-50, then both logistic and linear regression will provide similar results. The independent variables used in regression can be either continuous or dichotomous. In regression analyses, independent variables with more than two levels can also be used, but their conversion into variables is required to be done first that have only two levels. This is referred to as dummy coding. Generally, regression analysis is used with naturally-occurring variables, as opposed to experimentally manipulated variables, although one can use regression with experimentally manipulated variables. Causal relationships among the variables cannot be determined with regression analysis.
Simple linear regression is used when we want to predict values of one variable, when a value of another variable is known. The aim of regression analysis is to derive an equation of a line that aligns through that cluster of points with the negligible amount of deviations from the line. Standard multiple regression is the same concept as simple linear regression, but now we would be having several independent variables predicting the dependent variable. The significance levels assumed for each independent variable specifies whether that particular independent variable is a substantial predictor of the dependent variable, on top of the other independent variables. Because of this, an independent variable that is a significant predictor of a dependent variable in simple linear regression may not be significant in multiple regression.