Multiple regression (with two independent variables) with GraphPad Prism

GraphPad Prism is not designed for multiple variable analyses, but it can be configured to perform multiple linear regression with two independent variables. This example looks at how inflation relates to unemployment and percapita gdp. With most programs, you'd place each observation on its own row, with three columns for the three variables. With Prism, you enter data uniquely.

The dependent variable (inflation rate for this example) is entered as Y values running diagonally down the page. One independent variable (unemployment rate for this example) is entered into the X column. The other independent variable (per capita GNP for this example is entered into column titles. 

To fit the multiple regression model, you'll need to use a user-defined model. The easiest way to do so is to download and open this example Prism file, go to the parameters dialog for nonlinear regresion and click OK. Now the multiple regression model will be added to your list of user-defined equations. Alternative, enter this model as a user defined equation:

When defining the equation, set all three parameters to be shared (global fitting) and tell Prism that X2 is a data set constant equal to the column title. 

Also define the initial values for the three parameters to equal 0.0. Since this is a linear model, the initial values don't really matter. 

When you perform the nonlinear regression, check the option on the Diagnostics tab to report the adjusted R², as this is commonly reported with multiple regression. 

All the results you care about will be in the last, "Global" column of results. You'll see the best-fit values of the three parameters, along with their standard errors and confidence intervals.