poolRandomEffects
poolRandomEffects(
groups,level?):PooledEstimate
Defined in: random-effects.ts:50
Pool estimates from several groups under a random-effects model, by the DerSimonian–Laird method.
Use this when each group is one draw from a population of contexts rather than a repeat of the same measurement: separate processes, machines or sessions. A fixed-effect average would treat the between-group spread as if it did not exist and report an interval far too narrow; this estimates that spread as τ² and widens the interval by it.
With one group there is no between-group spread to estimate, and the result is that group’s own estimate and interval.
O(k) in the number of groups.
Parameters
Section titled “Parameters”groups
Section titled “groups”readonly RandomEffectsGroup[]
One entry per group. Not modified.
level?
Section titled “level?”number = 0.95
Confidence level in (0, 1). Default 0.95.
Returns
Section titled “Returns”The pooled estimate, its interval, and how much the groups disagreed.
Throws
Section titled “Throws”When groups is empty, when any variance is not a
positive finite number, or when level is outside (0, 1).