Scientific INDIA
C. R. Rao: the statistician behind modern data science
C. R. Rao (1908–2003) was an Indian mathematician and statistician whose work transformed how researchers design experiments, analyse uncertainty and draw conclusions from data. His career linked rigorous theory with national development, while his methods now influence fields ranging from clinical research to machine learning.
The term “data science” did not exist during Rao’s lifetime, but the principles behind it are recognisable in his research: identify the question carefully, collect informative observations, quantify uncertainty and avoid claims that the evidence cannot support. His life offers both an account of modern Indian statistics and a practical foundation for evidence-based work in Australia.
From Kanara to Cambridge
Rao was born in Hoovina Hadagali, in the Kanara region of present-day Karnataka, into a family of eight children. His father was a mathematics teacher, and his early home environment encouraged sustained intellectual curiosity. After his father’s death, his mother moved the family to Bangalore so that the children could receive better schooling. Rao studied mathematics at Central College and began publishing statistical papers while still an undergraduate.
| Statistical concern | Rao’s contribution | Modern data-science application |
|---|---|---|
| Experiments involving many variables | Orthogonal arrays and fractional factorial designs | Smaller, balanced experiments and faster factor testing |
| Limits on measurement precision | Cramér–Rao inequality | Bounds on the uncertainty of unbiased estimates |
| Efficient data reduction | Rao–Blackwell result | Removing noise while retaining sufficient information |
| Comparing statistical models | Rao score test | Efficient nested-model and diagnostic testing |
Rao moved to University College London in 1935, completed his doctorate and then worked at Cambridge’s Statistical Laboratory under R. A. Fisher. This period sharpened his interest in mathematical statistics, especially the design and analysis of experiments. The table shows that his achievements were not tied to any particular computer or software: they concerned the deeper question of how reliable conclusions could be obtained from limited and imperfect observations.
Building India’s statistical institution
Rao returned to India in 1939 and became a central figure in the creation of the Indian Statistical Institute under P. C. Mahalanobis. The institute offered an unusual combination of advanced research, teaching and practical statistical work. Rao helped establish it as an institution capable of supporting agriculture, industry, public health and national planning.
He also contributed to India’s system of official statistics, including work connected with sampling and the National Sample Survey. His approach linked mathematical theory to urgent national questions: how to estimate crop yields, population characteristics and economic conditions when complete enumeration was impractical. The resulting culture of organised data collection became part of the infrastructure of the independent Indian state.
Methods that shaped modern data science
Rao’s work on orthogonal arrays and fractional replication addressed a central problem in experimental science: how to investigate numerous factors without conducting an unmanageably large number of trials. These methods preserve useful information while reducing the number of combinations tested. Their descendants can be seen in modern A/B testing, industrial process optimisation and algorithmic feature evaluation.
His Cramér–Rao inequality established a lower bound on the variance of unbiased estimators under standard regularity conditions. In practical terms, it asks whether a data collection process is capable of producing estimates more precise than a theoretical limit. His score test, now commonly called the Rao test, compares a fitted model with a more restricted alternative by examining whether the available information contradicts the restrictions. Together, these ideas support uncertainty quantification, model selection and reproducible analysis.
The Rao–Blackwell theorem demonstrated that conditioning an estimator on a sufficient statistic cannot increase its variance. This principle remains valuable whenever a dataset contains repeated measurements, redundant variables or opportunities to summarise observations without losing decision-relevant information. Although his methods are usually presented through equations, they express a simple conviction: statistical analysis should extract signal while minimising avoidable noise.
A career crossing continents
Rao maintained an extensive international academic career, holding positions in India, Britain and the United States, including appointments at the University of Texas at Austin and the University of Pittsburgh. His books, especially Linear Statistical Inference and Its Applications, became standard references for generations of statisticians. He continued doing original research into his eighties, contributing to multivariate analysis, optimisation and the geometry of statistical models.
His honours included election as a Fellow of the Royal Society and the International Prize in Statistics. He also received India’s Padma Vibhushan. These achievements reflected a career devoted largely to institutions, textbooks and mentorship rather than commercial ventures. The lasting result is a body of methods taught across borders and applied wherever evidence must be converted into a defensible decision.
Rao’s legacy in an Australian context
Rao’s emphasis on sampling and transparent estimation remains particularly relevant in Canberra, where the Australian Bureau of Statistics produces census, labour-market and health datasets used across government and research. The Australian Institute of Health and Welfare, CSIRO and national universities similarly depend on statistical standards that make comparisons meaningful. His legacy is therefore visible less as a monument than as a habit of measuring carefully.
Australian data practice also reflects the country’s diverse markets and customs. Sydney’s growing fintech, health-tech and e-commerce sectors use experiments, prediction and uncertainty estimates to guide decisions. Melbourne’s sporting culture, including Australian rules football analysis and the predictive tradition surrounding the Melbourne Cup, demonstrates the demand for models that can explain uncertain outcomes. In Brisbane, data scientists work on challenges involving agriculture, resources and climate. Across these sectors, Australian employers value statistical judgement as well as programming ability.
Practical principles for evidence-based analysis
Rao’s work can be translated into guidance for analysts working with contemporary data and machine-learning systems:
- Design the study before collecting the data, including the comparison groups, time periods and decision thresholds.
- Match the method to the question, using prediction for forecasting and inference when uncertainty must be quantified.
- Report confidence intervals and practical error sizes rather than relying on significance labels alone.
- Remove redundant measurements where a sufficient summary preserves the information needed for the decision.
- Test assumptions about distributions, missing values, independence and model structure.
- Preserve reproducibility by documenting data versions, analytical choices and sensitivity checks.
For an Australian analyst in Canberra, Melbourne, Sydney or Brisbane, the durable lesson from Rao’s life is that credible data science depends on explicit design, honest uncertainty and disciplined claims.