How to Select a Data Analysis Technique

By Shane Hall

Choosing an appropriate statistical analysis technique for a given study is a difficult matter, affected by various considerations regarding the type of study being conducted and the nature of the data. Further, there is no single correct answer. Two different analysts with the same set of data could arrive at two different methods of analysis, both of which would reveal legitimate results.

Things You'll Need

  • Statistical or spreadsheet software
  • Set of data
  • Statistics book or guide

Step 1

Consider the type of study you are conducting and the key research question or questions. For example, a study that seeks merely to describe the characteristics of a given group would require a different analysis than a study that is interested in the effects of a given intervention, program or treatment. The former could be done with descriptive statistics, such as frequencies and averages. The latter, however, would require more complex analytical techniques, ranging from difference of means to regression analysis.

Step 2

Examine the data before settling on a specific analytical method. The nature of the data affect, and may even limit, the type of analysis you can conduct. Consider both the dependent variable (also known as the outcome variable) and the independent variable(s). Are these data categorical, ordinal or interval? Categorical data, in which the values of the data represent categories (such as gender, ethnicity, political affiliation), are the simplest. Attitude scales such as the Likert scale are examples of ordinal data. A Likert-type scale asks you to answer a question by making a choice along a range of answers numbered from, say, 1 through 5. With interval level data, the values tell you something about the difference between given values. Test score data, for example, are often interval.

Step 3

Consider the design of your study. The gold standard in most research is the randomized experiment, in which subjects are randomly assigned to the experimental and control groups. Many clinical trials of medications use this design. For various practical and ethical reasons, most social science research uses quasi-experimental research designs, in which the experimental and control groups differ on key characteristics. This is an important consideration for the analyst, who must control for these differences when analyzing the data. Often, quasi-experimental studies require more complex analytical techniques.

Step 4

Consider the capabilities and limitations of the software with which you will conduct your analysis. Some spreadsheet programs, such as Excel, can conduct most statistical analysis techniques, with the exception of some advanced multivariate procedures.

Step 5

Open Excel and click the "Tools" menu. Then select Add-Ins and check the "Analysis ToolPak." This will enable Excel to conduct data analyses. Users who want to expand Excel's analytical capabilities to include more advanced techniques can do so with a program known as XLStat, a product of Addinsoft, a software company specializing in analytics. XLStat enables Excel to conduct factor analyses, path modeling and other highly advanced data analysis procedures. XLStat is available for purchase and download from Addinsoft's XLStat website. In addition, a 30-day trial version is available for free.

Step 6

Consult your statistics book or guide to determine an appropriate analytical technique after considering your key research question, the nature of your data, the design of your study and the capabilities of your software.

Tips & Warnings

  • If you plan to conduct many advanced data analyses using large sets of data, consider purchasing a specialized statistical software package. The best-known statistical software programs used by scientists, statisticians and university researchers include SAS, SPSS and Stata.