lmsqreg.fit: Fit A Quantile Regression

Description Usage Arguments Value Version information: References See Also Examples

Description

Cole and Green LMS (Stat in Med, 1992)

Usage

1
2
3
4
5
lmsqreg.fit(YY, TT, edf = c(3, 5, 3), targlen = 50, targetx = seq(min(TT), max(TT),
        length = targlen), pvec = c(0.05, 0.1, 0.25, 0.5, 0.75, 0.9, 0.95), 
        maxit = 15, tol = 0.01, verb = FALSE, lam.fixed = NULL, mu.fixed = NULL, 
        sig.fixed = NULL, xcuts = quantile(TT, c(0.2, 0.4, 0.6, 0.8)),
sig.init, lam.init)

Arguments

YY

Ordinate values; must be positive

TT

Abscissa values

edf

A 3-vector specifying "equivalent degrees of freedom" for Box-Cox transformation, median, and standard deviation functions respectively, assumed to be smooth in TT. The scale of edf corresponds to the df parameter of s() in gam(), for which df=1 corresponds to a linear model. Constant models for the component functions can be obtained by setting lam.fixed, etc. to specified values (see below).

targlen

Number of points at which smooth estimates of L, M, S should be extracted for quantile plotting; if quantile plots are jagged in appearance, the value of this parameter should be increased. This parameter has no effect on the fitting process.

targetx

Points on which smooth estimates of L, M, S should be extracted for quantile plotting

pvec

Vector of target percentiles for plotting; default (5,10,25,50,75,90,95) percentiles will be plotted.

maxit

Limit to number of Fisher scoring iterations

tol

Tolerance on change between estimates of L, M, S on successive scoring iterations

verb

verbose run; will give iteration-specific information if T

lam.fixed

if NULL, lambda will be estimated; if non-null (numeric atom), lambda will be set to this value. Set to zero to force a log transformation; set to unity to perform no transformation.

mu.fixed

if NULL, mu will be estimated; if non-null (numeric atom), mu will be set to this value.

sig.fixed

if NULL, sigma (coefficient of variation function) will be estimated; if non-null (numeric atom), sigma will be set to this value.

xcuts

vector of x values defining classes within which Kolmogorov Smirnov (KS) tests for normality of derived Z-scores will be conducted.

sig.init

initialization

lam.init

initialization

Value

A list of class "lmsqreg.fit"; see lmsqreg.object for details.

Version information:

Document version 2.4 97/03/26 /usr16/stdevs/stdev0f/SLIBS/lmsqreg.dev.obs/SCCS/s.lmsqreg.fit.d

References

TJ Cole, PJ Green, Smoothing reference centile curves: The LMS method and penalized likelihood, Statistics in Medicine, v11, 1992, p1305–1319.

See Also

lmsqreg.object, print.lmsqreg.fit, plot.lmsqreg.fit

Examples

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
 set.seed(123)
 nnn <- runif(300, 10, 20)
 jjj <- 8 + 2 * sin(nnn) + rnorm(100, 0, nnn/11)
 fff <- lmsqreg.fit(jjj, nnn)
 fff      
# results from SPLUS:
#lms quantile regression, version 2.3, fit date Sat May 25 20:44:57 EDT 1996
#
# 
#Dependent variable: jjj , independent variable: nnn 
#The fit converged with EDF=( 3,5,3 ), PL= 299.045 
#
#                                                             
#  nominal percentile 0.050 0.100 0.25 0.500 0.75 0.900 0.950
#estimated percentile 0.053 0.113 0.25 0.503 0.73 0.887 0.947
#
# 
#Shapiro Wilk tests: (intervals in nnn //p-values)
#  9.999+ thru 11.974 11.974+ thru 14.257 14.257+ thru 16.251 
#               0.239               0.334               0.568
# 16.251+ thru 17.915 17.915+ thru 19.965 
#               0.174               0.191
#

lmsqreg documentation built on May 2, 2019, 6:47 p.m.

Related to lmsqreg.fit in lmsqreg...