A duration with timing variability: a log-normal draw whose mean is the
planned T and whose standard deviation is Wing and Kristofferson's
two-level sigma = sqrt(motor^2 + (CV T)^2). A duration is positive and
its spread grows with its length (the scalar property of timing, Gibbon
1977, Psychological Review 84:279), so its distribution is the
log-normal, the one of a positive quantity with multiplicative noise,
rather than a normal whose tail crosses zero: with
s^2 = ln(1 + sigma^2 / T^2) and m = ln T - s^2 / 2, the draw is
exp(m + s N(0, 1)), which has the two-level model's mean and variance
exactly and needs no floor.
A duration with timing variability: a log-normal draw whose mean is the planned
Tand whose standard deviation is Wing and Kristofferson's two-levelsigma = sqrt(motor^2 + (CV T)^2). A duration is positive and its spread grows with its length (the scalar property of timing, Gibbon 1977, Psychological Review 84:279), so its distribution is the log-normal, the one of a positive quantity with multiplicative noise, rather than a normal whose tail crosses zero: withs^2 = ln(1 + sigma^2 / T^2)andm = ln T - s^2 / 2, the draw isexp(m + s N(0, 1)), which has the two-level model's mean and variance exactly and needs no floor.