lmsqreg.fit: Fit A Quantile Regression
In lmsqreg: LMS (Cole-Green Stat in Med 1992 quantile regression)

Description Usage Arguments Value Version information: References See Also Examples

Cole and Green LMS (Stat in Med, 1992)

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)

`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

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

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

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

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

 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
#