Hello,
I am planning on doing a configurational analysis using either QCA or CNA on a small subgroup (n=15) of incarcerated individuals who do not report interest in release or reintegrating the community. They are part of a larger sample (n>500) of inmates.
To conduct the analysis in a valid way, do you recommend using all of the sample or a smaller random selection (say, 50 randomly picked inmates)?
Thank you in advance,
David J
On case selection for a configurational analysis using QCA or CNA with your subgroup of 15 incarcerated individuals lacking interest in release/reintegration, drawn from a larger sample (n>500). These methods excel in small- to intermediate-N designs (typically 10–50 cases) to identify causal configurations, requiring variation in both conditions and the outcome for valid inference on necessity and sufficiency.
Analyzing only the subgroup (all outcome-positive) would restrict you to necessary conditions, as outcome absence is needed for contrasting pathways. Using the full sample (n>500) risks excessive configurations and loss of interpretability, given exponential growth with conditions (2^k). A random subset (e.g., n=50) may imbalance outcome distribution, potentially underrepresenting your subgroup.
Instead, adopt a purposive approach for balance and limited diversity:
- Include all 15 outcome-positive cases.
- Select 15–35 outcome-negative cases purposively (e.g., matched on demographics or stratified for condition diversity) to reach N≈30–50, ensuring at least 2^k observations.
- Badges
- Science topic