Autonom: a shiny application for automated predictive analytics and nomogram visualisation

The rapid diffusion of data-driven decision-making has created demand for analytical tools that hide programming complexity while retaining statistical rigour. We present AutoNom, an R Shiny application that automates the full predictive-modelling pipeline—data import, exploration, multi-family regression, backward feature selection, internal validation, calibration, nomogram construction, and power analysis—through an intuitive point-and-click interface. Eight model families are supported (linear, logistic, ordinal, Poisson, quantile, Cox proportional hazards, accelerated failure-time, and generalised least-squares), each fitted with the rms package’s regression engine (Harrell, 2022). A fast backward step-down procedure guided by Akaike information criterion (AIC) reduces predictors to a parsimonious subset, and resampling routines (10-fold cross-validation by default) provide optimism-corrected performance indices. In a classroom evaluation (n = 42 undergraduates) the median time to build, validate, and interpret a logistic-regression model fell from 45 minutes (scripted R) to 12 minutes with AutoNom; the System Usability Scale mean was 86/100 (SD = 6). The current version extends a prototype previously reported by Abdullah (2024) by adding calibration curves, power calculators, and effect size estimation. AutoNom therefore offers educators, clinicians, and applied researchers a reproducible, statistically sound environment for predictive analytics without coding.