Case-Ordered Predictor-Outcome Matrix

Description

A cross-case matrix that orders cases by an outcome and shows, for each, how strongly a set of candidate predictors was present, to see which antecedents go with better or worse outcomes.

Predictor-outcome matrices array cases on a main outcome or criterion variable and give data for each case on the antecedent variables thought to contribute most. The principle is explanatory: do these antecedents predict or account for the outcome? In the book's example the outcome was ease of early implementation and the predictors were seven 'preparedness' factors (commitment, understanding, resources, skills, training and others, some broken down by role), chosen because they were manageable in number, plausible conceptually and supported by prior studies and informants. The outcome was scaled from the case reports by counting the number and gravity of reported problems, ordering cases from 'smooth' to 'rough'. Predictors were scaled from each case's checklist matrix on an ordinal scale from 'not in place' to 'fully in place', with marks for whether the factor was a facilitator or a barrier and whether the field researcher judged it decisive; a summed preparedness score was added. The authors stress that converting qualitative data to scales needs explicit decision rules and that a second analyst should verify judgements. For UX it supports questions such as which onboarding conditions distinguish accounts that adopted smoothly from those that churned.

Attributes

Loss type
Latent
Data
Mixed
Effort
Days

Steps

  1. Choose the outcome and order cases on it using explicit decision rules (for example count and weight of reported problems).
  2. Select a manageable set of predictors on conceptual and empirical grounds.
  3. Scale each predictor per case, for example from a checklist matrix, and mark whether it acted as facilitator or barrier.
  4. Enter cases as rows in outcome order and predictors as columns; add summary scores if useful.
  5. Look for predictors that vary consistently with the outcome, then check the case text for decisive factors.
  6. Have a second analyst verify the ratings and test conclusions on further cases.

Sources

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

Also known as

  • Predictor-outcome matrix
  • Case-ordered descriptive meta-matrix

Categories

  • User Research
  • Qualitative Analysis & Synthesis