Chi-Square

Do different groups have different preferences or outcomes?

The chi-square test is a statistical method used to determine if there’s a reliable relationship between two categorical groups.

For instance, suppose we’re interested in whether adoption of the attribution measures marketing-sourced, marketing-influenced, or both differs depending on whether they practice Account-Based Marketing (ABM) or not. Because both the type of attribution measure (sourced, influence, both) and ABM practice (practices or does not practice) are categorical factors, we use a chi-square test to analyze their relationship.

The chi-square test begins by calculating what percentage of all marketers use sourced, influenced, or both.

In this example, we found the following:

	Both	Influenced	Sourced
Total using	57.3 %	20.6%	22.1 %

Then, it calculates the same percentage for those with an ABM practice versus those without an ABM practice.

	Both	Influenced	Sourced
Total using	57.3 %	20.6%	22.1 %
ABMers	58.1%	22.5%	19.4%
Non-ABMers	53.9%	11.7%	43.4%

The overall measure tells the test what to expect if there is no difference. Then it compares what is observed in the data to what is expected and calculates whether that difference is likely by chance or if instead it is likely because a real difference exists between those with an ABM practice and those without an ABM practice.

The chart above indicates where there are meaningful differences in how ABMers and non-ABMers use attribution metrics. The Chi-square statistical test (below) tells us whether any of the differences observed are statistically reliable. In this case, the p-value of p>.001 indicates that there are significant differences. In this case, a majority of the two groups are equally likely to use both. However, for the remainder, ABMers are more likely to use Influenced metrics and non-ABMers are more likely to use Sourced.