Statistics and Visualization for Data Analysis and Inference
A brief class on analyzing behavioral data with visualization, resampling, linear models and Bayesian methods, offering an alternative to significance testing.
Null-hypothesis significance testing has limitations, and this brief class sketches an alternative way of analyzing behavioral experiments, built up by starting with graphs and simulated chance. It moves on through parametric distributions and the linear model toward Bayesian models of data, and it uses simple and complex behavioral experiments and eye-tracking as running examples.
What You Will Achieve
- Create visual displays that clarify experimental results
- Use resampling to judge what could happen by chance
- Fit and interpret generalized linear models
- Explain the limits of null-hypothesis significance testing
Program Overview
Visualization and Resampling
The class starts by looking at data and then uses non-parametric methods to assess it.
- Univariate and multivariate displays
- Non-parametric tests
- Confidence bounds and effect size
Distributions and the Linear Model
Next come parametric descriptions of data and the models built on them.
- Binomial, Gaussian and Poisson forms
- Bayesian derivation of tests
- Logistic regression
- Multilevel models
Bayesian Models and Examples
The last part builds more complex models of data and works through examples.
- Models with complex task assumptions
- Basic techniques for Bayesian inference
- Worked examples
Course Completion
Job Outlook
These skills are useful in behavioral science research, cognitive science and any lab or analytics role that handles experimental data.
Good for: Researchers who analyze behavioral experiments and want a different framework for inference, including those who are unhappy with routine significance testing.
Keep in mind: It is a very short January class from 2009, and notes are available only for selected lectures. Treat it as a sketch and use the Gelman references for depth.
Bottom line: A concise, opinionated introduction to data analysis for experimenters, useful as a map for further reading.
Frequently Asked Questions
Which software does the class use?
It says familiarity with MATLAB, Octave or R will be useful. No exercises are published, so you can try the methods in whichever of these you prefer.
What books does it point to?
It lists Gelman and Hill on regression and multilevel models, and Gelman, Carlin, Stern and Rubin's Bayesian Data Analysis, as supplementary references.
How much material is available?
The schedule has five lectures on visualization, resampling, distributions, the linear model and examples, with notes for selected lectures. There are no assignments or videos listed.



There are no reviews yet.