Sketching a Regression with Interaction


Return to Multi-way Page 3

Use the following equation to plug in the possible values for toilet and education and sketch the four resulting points in a table.  The combinations include:

    NOTOILET             DLOWEDN

1.        0                                  0
2.        1                                  0
3.        0                                  1
4.        1                                  1

 

Plug each combination into the following equation:

WAZ = -0.958 - 0.791*(NOTOILET) - 0.524*(DLOWEDN) + 0.834*(NOTOILET*DLOWEDN)

 

Here are the results for each:

1.  WAZ= - 0.958 - 0.791 * (0) - 0.524 * (0) + 0.834 * (0) ----->  WAZ= - 0.958
2.  WAZ= - 0.958 - 0.791 * (1) - 0.524 * (0) + 0.834 * (0) ----->    WAZ= -1.749
3.  WAZ= - 0.958 - 0.791 * (0) - 0.524 * (1) + 0.834 * (0) ----->   WAZ= -1.482
4.  WAZ= - 0.958 - 0.791 * (1) - 0.524 * (1) + 0.834 * (1) ----->    WAZ= -1.439

Here is a graph of wt/age (y-axis) against access to toilets (yes/no) on the x-axis, for 2 categories of education.

 

wpe9.jpg (23365 bytes)

INTERPRETATION:

The graph shows the outcome of WAZ score for all combinations of education and sanitation.  This includes consideration for an interaction between education and sanitation.  There is clearly a greater effect when sanitation is improved among those with >primary school level of education, whereas there is not anything worth mention in the group for low education.  These results are clear because we used an interaction variable in the model, which works very similar to breaking into categories of education status for regression analysis as we did in the previous example with water and sanitation controlling for education status. 

Return to Top of Page