An additive model with independent button delays and zero covariance with other errors gave a mean maximum delay share of 70.5% for 23 fixed ds006720 duration covariance matrices

We built a tool to bound variance from independent button delays in fixed duration covariance matrices. The additive model sets covariance between button delays and all other errors to zero. It allows different press variances and any covariance for the other errors. We used 3,289 complete three-interval trials from 23 adults in OpenNeuro ds006720 version 1.0.0. We removed condition and run means in seconds. The mean maximum delay share was 70.5%. The mean minimum remainder share was 29.5%. Each share uses the sum of the three interval variances. These means give each person equal weight. Correlated press delays could explain all supplied covariance. In 500 pure-delay Gaussian samples, the true remainder share was zero. The estimated minimum remainder share averaged 9.0%. Thus, the data result describes fixed estimates. It is not a population confidence bound or a memory-variance estimate. The package contains the proof, source events, hashes, code, numerical certificates, results, and reproduction instructions.

covariance bounds · duration reproduction · button timing · independent errors · semidefinite optimization · fixed covariance · sampling uncertainty