BayesWealth
A test of the common assumption that people save more as they age — and a null result reported honestly when the data pointed at financial literacy instead.
PyMCArviZCausal DAGMCMCNetworkX
1,000
Survey records
1.00
r_hat
500
Recovery samples
What I built
Encoded the “age drives saving” hypothesis as a causal DAG and fit two linked Bayesian regressions in PyMC over 1,000 survey records, sampling with MCMC to clean convergence (r_hat 1.00 on every parameter).
Reported what the posterior actually said: financial literacy predicted saving strongly (coefficient 1.685, 89% HDI [1.615, 1.753]) while age's effect was indistinguishable from zero — a null result on the original hypothesis, stated as such rather than massaged.
Checked the machinery, not just the answer: simulated 500 samples with known true coefficients, refit the model, and confirmed every true value landed inside its 89% credible interval.