powersolve_general: General solver for power law curve

In OptHoldoutSize: Estimation of Optimal Size for a Holdout Set for Updating a Predictive Score

General solver for power law curve

Description

Find least-squares solution: MLE of (a,b,c) under model ⁠y_i = a x_i^-b + c + e_i⁠; e_i~N(0,y_var_i)

Try a range of starting values and refine estimate.

Slower than a single call to powersolve()

Usage

powersolve_general(x, y, y_var = rep(1, length(x)), ...)

Arguments

x	X values
y	Y values
y_var	Optional parameter giving sampling variance of each y value. Defaults to 1.
...	further parameters passed to optim. We suggest specifying lower and upper bounds for (a,b,c); e.g. lower=c(1,0,0),upper=c(10000,3,1)

Value

List (output from optim) containing MLE values of (a,b,c)

Examples

# Retrieval of original values
A_true=2000; B_true=1.5; C_true=0.3; sigma=0.002

X=1000*abs(rnorm(10000,mean=4))
Y=A_true*(X^(-B_true)) + C_true + rnorm(length(X),sd=sigma)

c(A_true,B_true,C_true)
powersolve_general(X[1:10],Y[1:10])$par
powersolve_general(X[1:100],Y[1:100])$par
powersolve_general(X[1:1000],Y[1:1000])$par

OptHoldoutSize documentation built on April 12, 2025, 1:34 a.m.