**Correlation
Matrix**

Although it might not be the best way to see the whole picture, correlation matrices can be somewhat insightful... and easy to run. Try the following exercise to create a correlation matrix using the variables: WAZ score, education (dummy for low education), housing (dummy for bad roof), sanitation (dummy for no toilet access), water source (dummies for piped, well and river water), and income (virtual variable for income).

1. Open

keast4j.sav2. Click on

Statistics, Correlate, Bivariate.3. Enter the list of variables (

waz, dlowedn, dbadro, notoilet, dpiped, dwell, driver, vincome) one-by-one into the variables box using the arrow key.4. Click on the box to highlight

Pearson's Correlation Coefficientand the box forTwo-tailed.5. Click on

OK.

**INTERPRETATION:**

Correlations measure how variables or rank orders are related. The Pearson's correlation coefficient measures for a linear relationship between the variables of interest and it shows positive or negative direction. Significance is reported for two levels, * is at the 0.05 level and ** is at the 0.01 level. It is clear that there are strong correlations between the variables of interest and the outcome variable (waz score) for all but pump/tap water. There are also many of the independent variables that are significantly correlated with other independent variables such as bad roof with low education and income. In fact, most of these variables are correlated with one another. This collinearity between the variables must be given some attention in the analysis. As previously mentioned, don't let this worry you too much and don't fish for associations. For example, the associations between the variables are likely to be in the direction you would expect. Those with poor housing have less access to sanitation or latrines and a lower level of education. This also indicates that it is necessary to control for the overlapping effects from some of the independent variables.