Data can be analyzed in numerous ways. There are a few go-to methods that every researcher needs to be proficient with. As you create surveys for attendees and analyze the data, it's important to understand. The tool of choice for diving into survey data is the crosstabulation. There are fancier multivariate techniques, and those have their place, but for everyday use, the crosstab is the preferred method for analyzing nominal and ordinal data. Questions that generate these data types dominate most consumer and B2B market research surveys.
The crosstab (xtab for short) can accommodate two, or more, variables. Its purpose is to examine the shared distributions of the variables. When coupled with a statistical measure, such as the chi-square, the researcher can assess the degree of association between variables. Please note that I did not say causation, but association, this important distinction is reserved for another discussion.
The sample below provides an illustration of a two variable xtab with five levels for the row variable and two for the column variable (3 x 2). A third variable can be added to ‘control’ for potential influence. In this case, we could add gender to see if the relationship between job satisfaction and feelings of compensation is impacted by gender. We could add other variables such as age or income to further test the relationships in the data.
We view crosstabs from the perspective of rows and columns. For example, we can say that 57.8% of those who are satisfied in their current position feel they are paid fairly for the work they do. This compares to 24.4% of those who are not satisfied with their current role. In other words, those who feel satisfied with their current position are more than twice as likely to report they are paid equitably. From the column perspective, 48.8% of those who consider themselves unfairly paid are somewhat satisfied.
When making comparisons we can compare one cell to either another cell or to the row or column totals. For example, 59.6% of the fairly paid group were satisfied compared to 32.0% of the unfairly paid group. In total 43.7% of all respondents were satisfied. From the compensation perspective, 42.3% were fairly paid. The story we tell depends on a few aspects. The first thing to look at is whether or not the value of the chi-square is significant. If it is large enough for the significance to be at or less than .05 then we have a story to tell. In a table that is significant, you will see larger than expected differences in the percentages within a cell. If there is not significance then the cell percentages will be closer to their column and row total counterparts. If you are working with large tables (large as in more than 500 respondents) then you will want to see significance values closer to .000. The chi-square value is impacted by the number of respondents – it will grow as the response pool grows.
Lastly, using the table above, if our story focuses on satisfaction then we would read across the rows (% within Satisfied at work). If our focus is on compensation then we would read down the columns (% within Compensation). Take the time to get to know your rows and columns and you will become adept at converting data into insights. Need to know more about survey tools? Check out Cvent's survey solutions.
By Greg Timpany