Regression Stability Testing - Part 4

Scene 1:

Hello and welcome to the fourth video of our Regression Analysis Tutorial series. To watch the previous video click the annotation or the link in the description box. In this tutorial we'll focus on the advanced technique called regression stability test or Chow test to help us detect our model's deficiencies and the forecast reliability.

We'll use the same sample data as in our previous videos. It has been gathered from 20 different salespeople. Weekly sales is a dependent variable and the two explanatory variables are intelligence and extraversion.

Like before, take time to note how we organize our input data. Values for each variable are placed in separate columns and each observation or salesman is represented by separate row.

Next we'll again introduce the mask which is a boolean array that chooses which variable is included or excluded from the analysis. Let's use the results from the last tutorial and set the mask entry for intelligence to be 0 and for extraversion to be 1. Let's also exclude the observation number 16 because as we saw earlier it proved to be an influential data point in our regression model.

In order to examine the stability of our regression model we'll need to split the data into two non overlapping datasets, dataset 1 and dataset 2. For our example we'll pick the first 10 observations for dataset 1 and the remaining for dataset 2. The line outlines this operation.

The Chow test is going to construct three different regression models and attempt to find out whether any of the regression coefficients for either data set are significantly different from each other or from the combined dataset.

Now we're ready to begin! Select an empty cell in your worksheet where you wish for the output to be generated, then find the statistical test icon in the NumXL tab and click on the regression stability test or Chow test option.

Scene 2:

The regression stability test wizard pops up. Remember that by default the output cells range is set to the current selected cell in your worksheet. Now let's select the input cells range for dataset 1 and dataset 2, and specify the input mask. Once the data is selected the options and missing values tabs become available. For both let's select the default options.

Scene 3:

As a result of the automatic selection in the missing values tab, any missing value found in any of the input variables would exclude the observation from the analysis. Now click OK to generate the output table.

Scene 4:

The Chow tests results are generated. In our case the test accepts the null hypothesis. This means that the values of the coefficients are statistically indifferent and the entire dataset.

Scene 5:

That is it for now, thank you for watching!