criterium.stats.t-digest.scale — org.hugoduncan/criterium.arg-gen 0.5.245-ALPHA

Scale functions for t-digest algorithm. These control how cluster sizes are determined and affect accuracy in different ways.

bound clj

(bound x)

finite? clj

(finite? x)

k clj

(k scale q normalizer)

(k scale q compression n)

Convert quantile q to k-scale value using normalized compression.

k0 clj

Scale function that generates uniform cluster sizes. Used mainly for testing and comparison.

k1 clj

Scale function that generates cluster sizes proportional to sqrt(q*(1-q)). Gives constant relative accuracy if accuracy is proportional to squared cluster size.

k1-k-limit-f-low clj

k1-q-limit-f-high clj

k1-q-limit-high clj

k1-q-limit-low clj

k1-x-limit-high clj

k1-x-limit-low clj

k2 clj

Scale function that generates cluster sizes proportional to q*(1-q). Makes tail error bounds tighter than K1.

limit clj

(limit x low high)

max-size clj

(max-size scale q normalizer)

(max-size scale q compression n)

Maximum allowed cluster size at quantile q using normalized compression.

normalizer clj

(normalizer scale compression n)

q clj

(q scale k normalizer)

(q scale k compression n)

Convert k-scale value back to quantile q using normalized compression.