Academic journal article Reference & User Services Quarterly

# Statistical Significance

Academic journal article Reference & User Services Quarterly

# Statistical Significance

## Article excerpt

How It Signifies in Statistics Reference

Statistics-related reference questions may spawn heroic searches drawing upon reference librarians' ingenuity, knowledge of sources, and time. However, were librarians to understand better what many data-seeking users intend (or are assigned) to do with the numbers once acquired, they could communicate more effectively with users, find better-fitting data, and ration the heroics for "must" situations. Testing statistical "significance" is one of the major activities data-seekers perform. This article equips readers with enough of the basic meaning and method of statistical significance to improve their reference performance. It works through a two-category example using the t-test and a three-category example using analysis of variance (ANOVA), employing the widely available Statistical Package for the Social Sciences (SPSS). Several practical approaches are recommended for reference librarians to use when working with statistics reference questions.

A significant difference" ... "within the margin of error" ... "normally distributed"--such phrases all pertain to statistical significance. Reference librarians encounter them in graduate research-methods courses. Although some librarians continue to use them in their careers, and although most citizens hear them mentioned repeatedly in the popular media, the concepts behind statistical significance are, to say the least, imperfectly understood by most. Even so, statistical significance bears importantly on daily reference work in a way that reference librarians may insufficiently recognize.

This article contends that a better understanding of statistical significance is important for reference librarians. Reference desks tend to see a heavy traffic in statistics-type reference questions, and reference librarians are skilled--even heroic--at the retrieval entailed. That is not here challenged. However, knowing something about what users do with the data--after they gather it--has applicability to reference service. This article focuses on statistical significance. Another recent article in RUSQ likewise examined reference applications of correlation and regression analysis.(1)

By means of a desktop statistical package, users can readily check the statistical significance of their data after leaving the reference desk.(2) With today's plug-in-the-numbers software, statistical analysis has become easier and quicker than it was with yesterday's formulas and calculators. Reference librarians can learn the basics. Why specifically should they?

Why Reference Librarians Need to Know

There are three principal reasons why reference librarians need at least a grasp of the fundamentals of statistical significance. The reasons are outlined in brief below, then revisited after an example in which statistical significance testing is performed.

First, statistics-seekers, when collecting data, often announce at the reference desk that they are required to find at least a minimum number of observations or cases of the data. This is a direct consequence of the way significance is calculated, and reference librarians who understand that methodological constraint are likely to be better motivated and not to assume that users are at such times acting from quixotic notions or arbitrary impulses.

A second reason involves data aggregation (or disaggregation). Reference librarians know from experience that statistics found can vary considerably from statistics initially sought. Though unsuccessful at first, reference librarians often keep trying, perhaps taking heroic and inordinately time-consuming measures in pursuit of the data precisely as requested. However, knowing something about how significance tests work could lead to alternative approaches. The example that follows demonstrates how the "t-test," a commonly used test for measuring significant difference, compares data only in categories. …

Search by...
Show...

### Oops!

An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.