Epi Info™ User Guide

Classic Analysis


How to Use Advanced Statistics

Use the REGRESS Command

The REGRESS command performs linear regression and contains support for automatic dummy variables and multiple interactions.

REGRESS can be used for simple linear regression (only one independent variable), for multiple linear regression (more than one independent variable), and for quantifying the relationship between two continuous variables (correlation). Regression is used when you want to predict one dependent variable from one or more independent variables.

Syntax

REGRESS <dependent variable> = <independent variable(s)> [NOINTERCEPT] 
[OUTTABLE=<tablename>] [WEIGHTVAR=<weight variable>] [PVALUE=<PValue>]

Dialog Box


Figure 9.62: Linear Regression Window

 

  • The Outcome Variable is the dependent variable for the regression.
  • Other Variables appear in the predictor variables list.
  • Interaction Terms are defined with the Make Interaction button. Make Interaction appears if two or more variables are selected from the Other Variables list box. If you click Make Interaction, the relationship populates the Interaction Terms list box.
  • Make Dummy is activated if you select a predictor variable. If you select a numeric variable, it will be treated as discrete rather than continuous; the variable is enclosed in parentheses to indicate that dummy variables are being created. If a predictor variable enclosed in parentheses is selected in the list, the Make Dummy button changes to Make Continuous. Selecting it results in the variable being treated as continuous. If you select more than one predictor variable, the Make Dummy button changes to Make Interaction. Selecting it results in all possible combinations of the selected variables being added to the regression as interaction terms.
  • A Weight variable may selected to use in weighted analyses.
  • Confidence Limits specifies the probability level at which confidence limits are computed (default=.05).
  • The Output to Table field identifies a table to receive output from the command. (Currently Disabled).
  • If you select No Intercept, the regression is performed without a constant term, forcing the regression line through the origin.
  • OK accepts the current settings and data, and subsequently closes the form or window.
  • Save Only saves the created code to the Program Editor, but does not run the code.
  • Cancel closes the dialog box without saving or executing a command.
  • Clear empties the fields so information can be re-entered.
  • Help opens the Help topic associated with the module being used. (Currently Disabled).

How to Use

  1. From the Analysis Command Tree, use the READ command to open a PRJ project file. Select a form or table.
  2. From the Analysis Command Tree, click Advanced Statistics > Linear Regression. The REGRESS dialog box opens.

    Figure 9.63: Linear Regression Window

     

  3. From the Outcome Variable drop-down list, select a variable to be the dependent variable for regression.
  4. From the Other Variables drop-down list, select the variable(s) to be the predictors.
    • If you select any predictor variables, Make Dummy will be activated. Selecting one will allow a numeric variable to become discrete rather than continuous; the variable is enclosed in parentheses to indicate that dummy variables are created. If a predictor variable enclosed in parentheses is selected, the Make Dummy button changes to Make Continuous. Selecting it makes the variable continuous. If you select more than one, the Make Dummy button changes to Make Interaction. Selecting it adds all possible selected combinations to the regression as interaction terms.
    • Make Interaction appears if you select two or more variables from the Other Variables list box. Interaction Terms are defined with the Make Interaction button. Click Make Interaction. The relationship populates the Interaction Terms list.
    • The Output to Table field identifies a data table to receive output from the command. Results can be sent to an output table for graphing. (Currently Disabled).
    • If No Intercept is selected, the regression is performed without a constant term, forcing the regression line through the origin.
  5. Click OK. Results appear in the Output window.

Try It

  1. Read in the project Sample.PRJ. Open BabyBloodPressure.
  2. Click Linear Regression. The REGRESS dialog box opens.
  3. From the Outcome Variable drop-down list, select SystolicBlood.
  4. From the Other Variables drop-down list, select AgeInDays.
  5. From the Other Variables drop-down list, select Birthweight.
  6. Click OK. Results appear in the Output window.

Top