TABLE 3.
Descriptive analysis methods used in Exemplar's research
| Purpose and methods | Summary |
|---|---|
| Equity analysis1 | |
| Disaggregation by subnational geography, wealth quintile, maternal education, urban vs rural residence, and child gender | Estimates of stunting prevalence for key subpopulation |
| SII and CIX | Measure absolute and relative wealth inequalities, respectively |
| CIX | Estimated from logistic regression models of the cumulative distribution of the asset index, plotted against stunting prevalence |
| 5 × 5–km geospatial estimates | Modeled estimates that use location data to estimate the subnational distribution of stunting prevalence at 5 × 5–km granularity |
| Rates of reduction2 | |
| CAGR | Assessed relative change (decline) in stunting prevalence over time for each geographic region |
| AAPC | Estimated through ordinary least square regression models; stunting prevalence regressed on survey year |
| Population shifts in growth faltering | |
| Victora curves | Smoothed local polynomial regressions to depict HAZ vs. child age (in months) predictions with 95% CIs, estimated by surveys (e.g., DHS or MICS) |
| HAZ kernel density plots | Depict the distribution of child HAZ scores and enable assessment of skewness and kurtosis; stratified by child age groups: <6, 6–23, or ≥24 mo |
1
Analysis accounts for survey design and weighting. AAPC, average annual percentage point change; CAGR, compound annual growth rate; CIX, concentration index; DHS, demographic and health surveys; HAZ, height-for-age z-score; MICS multiple indicator cluster surveys; SII, slope index of inequality.
2Analyses conducted to show changes over survey round.