Stata - 18 ^new^

One of the most exciting announcements in is the deeper integration with Python. Data scientists no longer have to choose between Stata’s ease of use and Python’s machine learning libraries.

For users working with billions of observations, Stata/MP 18 unlocks deeper multi-threading capabilities. Feature Area Speed Improvement (Stata 17 vs Stata 18) Core Optimization Type 2x – 4x Faster Parallel Radix Sort Collapsing ( collapse ) 1.5x – 3x Faster Optimized Multi-Threaded Hashing Reshaping ( reshape ) Up to 2x Faster Memory Mapping Redesign 🛠️ Integration with Python and R Stata 18

The cumulative effect of Stata 18’s new features is transformative for research workflows. The combination of framesets and alias variables fundamentally changes how analysts manage multiple datasets, eliminating many merge operations and reducing memory usage. The table creation enhancements dramatically reduce the time spent formatting results for publication, allowing researchers to focus on interpretation rather than output formatting. The integration with Python opens new possibilities for machine learning and custom visualizations while maintaining Stata’s statistical rigor. One of the most exciting announcements in is

For example, you can estimate county-level poverty rates from a state-level survey by borrowing strength from covariates like tax returns and demographic data. implements both direct and model-based estimators (Fay-Herriot, unit-level). Feature Area Speed Improvement (Stata 17 vs Stata

| Feature | Stata 17 | Stata 18 | |---------|----------|----------| | Bayesian multilevel | No | Yes ( bayes: meglm ) | | Heterogeneous DID | Limited | Full (Callaway-Sant’Anna, Sun-Abraham) | | Python integration | Basic (via python command) | Bidirectional with pandas/NumPy support | | Small-area estimation | User-written only | Native sae command | | PNG/PDF export to PowerPoint | No | Yes ( putpptx ) | | Stata Markdown | No | Yes ( .smd files) | | Caching for reproducibility | No | Yes (hash-based) | | Performance (large merges) | Baseline | ~25% faster |

The corrected AIC and consistent AIC calculations in Stata 18 address known issues with sample size. AICc, introduced in Stata 18, is considered a more reliable criterion than AIC for small-sample datasets, helping researchers select appropriate model complexity without overfitting.

For a full breakdown of every technical addition, you can explore the official New in Stata 18 feature list . New reporting features | New in Stata 18