Ten Basic Steps...

Win O R S, O-L-S Regression …

Step One: Start Win O R S.   Additionally, open a Microsoft Word document. If this is the first Word document for this assignment, then start Word by use of the Windows Start menu. If you have previously created a Word document then you must open the document by using the following Win O R S menu tree:. Solution.. Open M-S Word document…

Step Two.  Use the File / Open menu sequence to load the H W K B A N K file into the spreadsheet. Enter variable names along row one.  Enter a one letter variable type, that is -- a V-Type, across row two.  Possible entries include: I.  D.  F.  G.  or W.

Step Three.  Select the menu sequence:. Applications. Statistical Methods. Regression.  The Regression Methods dialog box appears next.  Set the input range.  Choose the left-facing red arrow to roll-up the dialog box. Highlight all data cells that may be considered for analysis.  For example, the range may contain all cells from B-four to A--E fifty--eight. The forecast range is not discussed here.  It is an advanced regression option.  Accept the default entry for the input box titled: Insert On range. In the block titled Method Choices, click on the radio button associated with Ordinary Least Squares, or O-L-S. 

Step Four. In the section-block titled Options, for the Intercept option, accept the default check box. Click the check-box for V-I-F, or Variance Inflation Factors.  Also, do the same for the  Whites' test. This is a test for constant variance.  Accept all default settings in the section titled: Setup Files For.

Step Five. To the far right, in the section: Data Types.  Click on the radio button that best identifies your data.  NOTE: When time-series is chosen, the Elasticity check-box is active.  Click this to the On state.  Click O-K to generate a regression solution.

Step Six.  Analyze the regression results.  First check the validity of the regression parameters.  When V-I-F reports a problem with multicollinearity drop or transform offending variables.  Upon the elimination of collinear variables go back to step number 5 and solve the revised model.

Step Seven.  Continue the analysis.   Time Series Models. Check the two-tailed Durbin-Watson test.  This is displayed on row twenty-six of the ANOVA tab. If your time-series model fails this test, you will see the word Reject displayed in cell B twenty-six. Immediately go to menu tree: Solution. Current Files. Statistical Methods. Regression Analysis. Open the adjustment file that was recommended on the ANOVA tab.  Upon opening this file, do not modify any settings.  Simply solve for the O-L-S estimates.  The data transform may, or may not, improve the results.

Step Eight:  All model builders must check for constant variance in the residual error term.  Interpret the White's test P -- Value. For example, it should be a number greater than -- 0.05. If the constant variance assumption is violated then go to the following menu tree: Solution.  Current Files. Statistical Methods. Regression Analysis. Open the file: Weighted Least Squares. This file contains the data of the current regression model with one appended column.  The added column is a weight that is derived by the square of the individual error terms.  This variable has a V-type of W.  This is a common adjustment that will assist in, but not guarantee, the removal of non-constant variation in the error term. After invoking any combination of the data transformations discussed above, the resultant model is suitable for the interpretation of causal effects only. That is, the model cannot be used for econometric forecasting purposes.

Step Nine: If you have completed your analysis, print the results to the attached Word document. then go to step ten. If you have additional exploratory regressions to execute, go back to step five…

Step Ten:  Save, edit, and close all documents and applications.

This completes the ten basic steps for O-L-S regression analysis.

[WinORS Basic Steps] [Finance] [Statistics] [OLS Regression] [Stepwise Regression] [Value-at-Risk fx] [Forecasting] [Learning] [Math Prog]

 NKD-Group, Inc. 1998 - 2004