10 things to know about quantitative data: A guide for museum practitioners
Recently, I was invited to speak about quantitative data for The George Washington University’s course, Evaluating Museum Learning. My task was not to teach students how to conduct quantitative analysis, but rather, to help students become knowledgeable interpreters of quantitative data—whether encountered in their future museum work or the daily newspaper. Based on my experience presenting quantitative results to museums, I came up with a list of 10 things to know about quantitative results so you can better assess their meaningfulness.
Sample size is the part of the population selected for study. For example, you will not survey every museum visitor (if you did, that would be a census), but a sample of the population. There are tons of sample size calculators out there that can help you determine an appropriate sample size.
A sampling protocol, or recruitment strategy, describes how a sample was obtained. It answers the questions: Who was approached? How were they approached?
Response rate, or participation rate, indicates the percentage of the people recruited for a study who agreed to participate. It can be an indicator of representation issues with your sample. See this former blog post about representation.
Statistical significance is a measure of whether your research findings are meaningful (or technically, whether you can reject the null hypothesis). Statistical significance is determined by looking at the p-value, or probability value. It is common to use p-values of 0.01 or 0.05, meaning results less than or equal to this p-value are considered statistically significant.
Normal distribution is a distribution that occurs naturally in many situations (for example, test scores are usually normally distributed, along a bell-shaped curve). Some statistical tests are appropriate for normally distributed data (called parametric tests), and some are not (called nonparametric tests).
Variable types for quantitative data can be numeric (e.g., age) or categorical (e.g., member versus non-member); this distinction is important to consider when designing questionnaires and contemplating which statistical tests you might want to conduct.
Analysis of variance (ANOVA) with the F-test is an inferential statistic run with numerical variables that compares the means of two or more groups (for normally distributed data). For example, you could compare the age of members and non-members.
Cross tabulation with the chi-square test is an inferential statistic run with categorical variables that compares the percentages of two or more groups (for normally distributed data). For example, you could compare the percent of members who live in or out of state.
There are numerous types of statistical procedures! You may encounter many other statistical tests, such as linear regression, K-means cluster analysis, hierarchical cluster analysis, Kruskal-Wallis, Mann-Whitney, etc. (the latter two being tests for data that are not normally distributed).
Distorted or misleading graphs are everywhere! Beware of a y-axis that doesn’t start at zero, misleading titles, missing sample size, etc. You can find some great examples here. Also, check out my graph below, an example of how Microsoft products will automatically distort your graph.
