Statistician's Marriage Proposal Includes Confidence Interval

A statistician proposed to his partner last Saturday using a presentation that included a confidence interval for the probability of a successful marriage, a decision that his partner has accepted with what she describes as "reservations about the methodology." Dr. Karl Estimator, 34, presented a twelve-slide deck to his girlfriend of four years, Dr. Lisa Heuristic, during what she had been told was a "casual dinner at home." The presentation, titled "A Bayesian Analysis of Long-Term Partnership Viability: Our Case Study," concluded with a proposal and a 95% confidence interval of [0.78, 0.94] for marital success. "The point estimate was 0.86, which is well above the population base rate of approximately 0.50," Estimator explained. "I thought she'd appreciate the rigor." Heuristic, who is herself a professor of statistics at Columbia, said yes but immediately raised methodological concerns. "His prior was informative, which I found presumptuous," Heuristic noted. "He used our arguments-per-month rate as a covariate but failed to weight by argument severity. He also excluded the incident in Lisbon, which I consider an outlier but which should at minimum appear in the sensitivity analysis." Estimator acknowledged the omission of the Lisbon incident, calling it "a judgment call about data quality that I stand by." The couple has agreed to marry in June, pending Heuristic's independent replication of the analysis. She has requested access to Estimator's full dataset, including variables he described as "supplementary" and she describes as "suspiciously omitted." "Love is not a point estimate," Heuristic reflected. "It's a distribution. And I would have preferred he use a 99% interval for something this important. But the effect size is significant and the p-value is small, so I said yes." Their wedding invitations will reportedly include the phrase "You are cordially invited to witness a statistically significant union."