Within-person data won’t always match data from large groups. That’s why you should use experience sampling.

By November 8, 2016Experience Sampling
photo-1478184096253-c2d96e9263c1

Synopsis of Ellen Hamaker’s chapter “Why Researchers Should Think ‘Within-Person’: A paradigmatic rationale” in Handbook of Research Methods for Studying Daily Life

For many years, psychological research analyzed large samples of people in order to describe the aggregate whole. The only problem is that what is true about the aggregate is often not applicable to the individual.

To illustrate this example, consider the relationship between typing speed and the percentage of typos made. If you were looking at the whole population, the typing speed and typos made would be negatively related. This is because as skill level increases, so does typing speed. Therefore, the faster people can type, the less likely they are to make a typo (since they are more experienced).

However, examining this same relationship on a within-person level produces the opposite relationship: as typing speed increases, the number of typos will also increase to form a positive relationship. This makes sense because if Sally was to type faster and faster, Sally would continue to make more mistakes.

But what does this mean? The assumptions made about the aggregate population through large sampling (often through cross-sectional research) cannot be applied to the within-person level. Therefore, an alternative research method is needed to study the individual.

This is why daily life measures are valuable on the within-person level.

Hamaker uses the rest of her chapter to discuss a brief history of large sampling, why large sampling approaches are limited, and some alternatives to large sampling.

  1. Large-sampling approaches have been widely used for psychological research for many years.

    With the technology available in the 1900s, the easiest way to research a topic was by surveying a large population. Furthermore, most of the research regarded real life problems which were most easily analyzed by surveying a large population. This resulted in a “triumph of the population” which contributed to the belief that only large sampling approaches could present psychologically relevant data. Nobody gave much thought to any alternative sampling methods; large scale population research was the only way to do it for many years. 

     

  2. Since within-person and between-person levels contribute to large sample results, it is erroneous to assume that large sample results are true for the within-person level.

    Ergodicity is defined as the situation when specific mathematical-statistical conditions are met. In order to assume that large sample results are true on a within-person level, the means, variances, and covariances of the population must match the within-person statistics. Hamaker states that these variables are “very unlikely to hold in psychological practice” and “most -if not all- psychological phenomena are nonergodic.” While Hamaker also presents numerous mathematical relationships and comparisons of various statistics, the same point is made: large sample results cannot be applied blindly to the within-person level. Both between-person variables and within-person variables influence the cross-sectional relationship in varying proportions. 
  3. Time series single-subject and multilevel approaches are alternatives to large sampling.

    The single-subject approach utilizes time series analysis, which studies the relationships between variables to themselves and to each other over time for one specific case. This single-subject approach can be used to analyze within-person variability for one individual and has been frequently used in previous studies. However, the relatively new multilevel approaches are also valuable. They make use of time series models and focus on both between-person and within-person differences. This enables multilevel modeling to measure data that contains a trend and for data that is stationary. Additionally, time series multilevel modeling can describe psychological processes better than just regular multilevel modeling.

Hamaker notes that within-person questions have been asked in psychology for years; however, they have always been incorrectly assessed by using population data… until now. Now that we have the statistical techniques and multilevel modeling necessary to study these questions, psychological change on a within-person level can be researched.

Hamaker, E. L. (2012). Why researchers should think “Within-Person”: A paradigmatic rationale. In M. R. Mehl & T. S. Conner (Eds.), Handbook of research methods for studying daily life (pp. 3-21). New York, NY: The Guilford Press.

Want to try out experience sampling to study within-person variables yourself?

Sign up for a Free Trial