**Conduct multiple regression**

This function is applicable to any multivariate dataset or two-way table that can be regarded as a multivariate dataset. Follow the following steps:

- Delete any unwanted independent variables using the function "
*Data/Run SUBSET by/Removing any Testers*"; - Under
*Multivariate*, clicking*Multiple Regression*, a panel will appear above the biplot, which allows you to select the dependent variable. - Select the dependent variable from the drop-down box and click
*Regress*. The program will try to conduct multiple regression using the selected variable as dependent variable and all others as independent variables. - If the number of independent variables are more than the number of observations, a message will appear, telling you that over-parametrization has occurred, and suggesting you to reduce the number of independent variables using "
*Find Associates Testers of*…" function to remove less associated independent variables. - When done, run
*Multivariate /Multiple Regression*again. Once the over-parametrization problem is solved, the following will appear on the biplot:

1. The dependent variable in a larger font and underlined;

2. The retained independent variables;

3. Variable vectors to facilitate visualization of the relationships among the variables, particularly those between the dependent variable and the independent variables. For interpretation;

4. Total variation of the dependent variable explained by the multiple regression, on the top of the biplot; and

5. Numerical results, including the partial regression coefficient, t-statistics, type I error probability of the partial regression coefficients, and simple r-square values (variation of the dependent variable explained by each independent variable). This information is also printed to the log file for your reference.

**Note**: The simple r-square values are more correct when the multiple regression was preceded by the execution of the "find associates" function).

Follow this link for an example: QTL identification based on multiple linear regression.