No matter if you’re running human behavior studies in academic or commercial research, you certainly wonder how to analyze all the data you have been collecting over the course of several weeks (or months) and get wooing results.
Ideally, you have drawn up a general research approach back when you designed your study, phrased your hypothesis and selected the sensors, stimuli and respondent groups of interest.
Here is an example: You want to find out which one of two chewing gum packages generates higher shopper interest and engagement. For that purpose, you decide to use eye tracking to monitor the number of fixations towards each of the presented chewing gum packages along with EEG to assess your respondents’ cortical engagement levels (based on frontal asymmetry, for example). Your assumption is that higher interest and engagement are reflected by a higher number of fixations and higher frontal asymmetry scores. To test your hypothesis statistically, you select a simple t-Test to compare the two packages statistically.
There’s obviously dozens of statistical software solutions available to analyze your data, however only a few live up to the standards of human behavior research.
Basically, the software you choose for data analysis should allow you to do the following:
- Select respondents and groups for analysis. Maybe certain respondents had bad or missing data – these participants should be excluded or flagged so that they are not used for the analysis.
- Pre-process data to filter outliers, generate aggregates and averages over time, sensors, stimuli or respondent groups, etc.
- Compute descriptive statistics including measures of central tendency (mean, median) and dispersion (standard deviation, variance, range, etc.). These parameters are used later on for testing.
- Run statistical tests such as t-tests, ANOVAs, variance analyses, regressions and correlations as well as non-parametric procedures (bootstrapping, Monte-Carlo procedures).
- Plot data, for example histograms or scatter plots of the raw data distribution or statistical results.
We recommend to study online resources and compare the features offered by the different statistical packages.
To get a foot in the door and make the search even easier for you, we have put together a list with our top 4 candidates that will certainly help you get on track with your data analysis. Have a look below:
1. Microsoft Excel
Somewhat surprisingly, MS Excel brings a wide variety of tools for visualization and statistical analysis of your physiological data. Data import from text files is as simple as generating summary metrics and customizable graphics and figures.
What you get with Microsoft Excel:
- Excel offers a lot of control and flexibility.
- Excel is widely available and relatively inexpensive for students and private entities.
- Excel doesn’t require to learn new methods of manipulating data and drawing graphs.
- Results (data and figures) can easily be exported and imported for use in Microsoft Word and PowerPoint.
- Data manipulation in Excel is much more transparent compared to other tools.
- Excel offers a graphical user interface.
2. SPSS (IBM)
SPSS, (Statistical Package for the Social Sciences) is a generic, quite comprehensive analysis software, comprising descriptive statistics, parametric and non-parametric analysis functionality. SPSS plots are commonly found in academic papers and commercial research reports.
What you get with SPSS:
- SPSS has an efficient data management and offers a lot of control over data organization.
- SPSS offers a wide range of methods, graphs and charts.
- SPSS makes certain that the output is kept separate from the data itself, generating well-structured reports and worksheets containing results.
- SPSS has a scripting language that allows to generate scripts and templates for bulk processing of datasets and parameters.
- SPSS offers both a graphical user interface and a scripting interface.
Did you know? With a simple free add-in for Excel you can work similar magic as with SPSS.
3. MATLAB (The Mathworks)
MATLAB is a generic analysis framework, which requires programming skills to a much greater extent than Excel or SPSS. MATLAB contains an impressive collection of analytic libraries growing every day – even when a desired procedure is not available, you can script it yourself.
With MATLAB, you can clean up and pre-process your data, but also generate results and graphics, which you can fully customize. However, adapting figures for use in a publication or report can be quite challenging for beginners.
What you get with MATLAB:
- MATLAB offers specialized toolboxes for the analysis of data stemming from eye tracking, EEG, ECG, EMG etc. and facial expression analysis.
- In MATLAB, analytics, processing steps and outcomes can be completely customized.
- MATLAB offers academic licenses at a reduced price.
- Dependent on the toolboxes used, MATLAB offers graphical user interfaces, but you can also just work “in code”.
4. R (R Foundation for Statistical Computing)
Similar to MATLAB (but freely available), R offers all necessary data transformation and analysis tools a human behavior researcher can think of. R libraries are often considered to be somewhat richer than those of MATLAB.
What you get with R:
- R is completely free of charge.
- With R you can produce high-quality figures and plots.
- R has a large, active, steadily growing community of users, contributing to the active development of toolboxes and extensions.
- R scripts can make use of multicore processing, allowing you to get results faster compared to other tools.
Reach out to our experts at iMotions to learn more about statistical analysis software packages. We’re happy to help!