Reading focus
Syllabus: How Do Close Relationships Form and Function?
Distinguish predicting satisfaction levels from predicting change.
Why you’re reading it. Assess how much a broad set of baseline measures can establish about relationship trajectories.
A question to consider. Why could a measure predict later satisfaction yet offer little help in identifying whose satisfaction will rise or fall?
Why this reading
Predicting who reports a satisfying relationship differs from predicting whose satisfaction will change. Combining 43 datasets makes that distinction visible and tests the added value of different kinds of information.
Before you read
A random forest combines many decision trees. Held-out predictions use observations not used to fit a particular tree. Baseline level, later level, and change are distinct outcomes. Predictive performance does not identify an effective intervention.
Read for
- Compare relationship-specific reports, individual characteristics, and partner reports.
- Distinguish prediction within each dataset from combining performance across datasets.
- Compare performance for satisfaction levels with performance for change.
Consider afterwards
- If appreciation predicts satisfaction, what additional evidence would justify an intervention intended to increase satisfaction by changing appreciation?
Editorial context
The 43 studies include 11,196 couples from six Western countries and heterogeneous samples, including a large low-income sample. Most predictors are self-reports. Limited added prediction from partner reports does not mean partners are unimportant. Poor prediction of change from these baseline measures does not show that change is inherently unpredictable.