Post Three: Results

Hey kids! See if you can interpret a statistical regression using the data below!

  Coefficient P-value
Union Membership -3.36 0.003
Mexican Border 0.467 0.197
Perot Vote 0.565 0.574
Household Income 0.033 0.076
Corporate Contributions -0.313 0.604
Labor Contributions -4.137 0.000

Above are the results of the ordered probit regression I conducted over the summer. This sort of regression is used when the independent variable is “ordinal.” An ordinal variable is one where the different values have a clear order (such as high to low, or big to small), but the values are not distributed in evenly spaced intervals. (For example, “level of education” is ordinal, because “Master’s degree” is higher than “secondary education.” However, “years of education” is not ordinal, because you could put the possible values on a number line with easily quantifiable distances between the values.)

My independent variable is the “shift” variable, which has the values “toward yes,” “no change,” and “toward no.” Since these values can be ordered, the variable is ordinal.

The two columns in the chart above allowed me to determine the “statistical significance” and the “substantive significance” of the variables. Substantive significance tells you whether an increase in one variable leads to an increase in the other. A good way to visualize substantive significance is the slope of the regression line– if the xy-graph between the two variables is a horizontal line, the variables are not substantively significant. But if the graph is steep, this suggests a correlation.

However, statistical significance is also necessary to tell you whether the relationship actually exists. In other words, do the data points closely hug the regression line? Or is the data so scattershot that our regression is meaningless?

A high coefficient value (positive or negative) suggests substantive significance, while a low P-value (usually below .005) suggests statistical significance.

Using the data above, which of the variables are significant? (Remember– it must be both statistically and substantively significant)

In my next post, I will give you the answer.