Qualitative Cross-Case Scatterplot

Description

Plot each case as a point on two carefully scaled qualitative dimensions to reveal clusters, distances and outliers that a matrix's cut points hide, then return to the case data.

Matrices sort cases into cells and throw away how close or far apart they are. The scatterplot keeps that information: the analyst scales each case carefully on two dimensions thought to be related, plots them, and reads the pattern against a stated hypothesis. Where a value changed between periods, two points joined by an arrow show the movement; over-time variants place cases in panels for successive periods and circle case families. Families then prompt a search for an intervening variable. In the source example, pressure to adopt plotted against latitude to change refuted a 'bargain' hypothesis and showed latitude rising later. Scaling may need summed indices; when cases cannot be scaled finely, the analyst falls back to a contingency table. Conclusions must be confirmed in the case reports, with attention to reversed or third-variable causation. Plotting 12 already scaled cases takes 10 to 15 minutes.

Attributes

Loss type
Latent
Effort
Hours

Steps

  1. Pick two dimensions you believe are related and state the expected pattern.
  2. Scale each case on both dimensions, building summed indices and returning to case reports where needed.
  3. Plot the cases, drawing arrows where values changed between periods.
  4. Read clusters, quadrants and outliers against the hypothesis.
  5. Circle case families and look for an intervening variable that explains them.
  6. Check outliers and conclusions in the case data, watching for reversed or third-variable causation.
  7. Fall back to a contingency table if the scaling is too coarse.

Sources

  • Miles, M. & Huberman, A. M., Qualitative Data Analysis: An Expanded Sourcebook, 2nd ed. (1994)

Also known as

  • Case scatterplot
  • Scatterplot over time

Categories

  • User Research
  • Qualitative Analysis & Synthesis