R help - Sep 2010 - How using the weights argument in nls2?

Good morning gentlemen!

How using a weighted model in nls2? Values with the nls are logical since
values with nls2 are not. I believe that this discrepancy is due to I did
not include the weights argument in nls2.

Here's an example:

MOISTURE <- c(28.41640,  28.47340,  29.05821,  28.52201,  30.92055, 
31.07901,  31.35840, 31.69617,  32.07168,  31.87296,  31.35525,  32.66118, 
33.23385,  32.72256,
32.57929,  32.12674,  52.35225,  52.77275,  64.90770,  64.90770,  85.23800,
84.43300,  68.96560,  68.41395,  70.82880,  71.18400,  96.13240,  96.07920,
95.35160,  94.71660,  87.59190,  88.63250,  89.78760,  90.17820, 88.46160,
87.10860, 94.86660,  94.51830,  75.79000,  76.98780, 144.70950, 143.88950,
111.58620, 112.71510, 120.85300, 121.43100, 116.34840, 114.87420, 195.35040,
191.36040, 265.35220, 267.25450, 227.13700, 228.78000, 238.37120, 242.70700,
299.54890, 291.04110, 220.09920, 219.82650, 236.79150, 243.70710, 208.79880,
208.12260, 417.21420, 429.59480, 360.91080, 371.66400, 357.72520, 360.53640,
383.82600, 383.82600, 434.02700, 432.57500, 440.56260, 438.32340, 468.69600,
469.82140, 497.93680, 497.17010)

YEARS <-  rep(c(86,109, 132, 158, 184, 220, 254, 276, 310,
337),c(8,8,8,8,8,8,8,8,8,8))

VARIANCE <- c(2.0879048 ,   2.0879048,    2.0879048,    2.0879048,   
2.0879048,
 2.0879048,    2.0879048,    2.0879048,    0.3442724,    0.3442724,
0.3442724,    0.3442724,    0.3442724,    0.3442724,   0.3442724,
0.3442724,  151.9481710,  151.9481710,  151.9481710,  151.9481710,
151.9481710,  151.9481710,  151.9481710,  151.9481710,  115.3208995,
115.3208995,  115.3208995,  115.3208995,  115.3208995,  115.3208995,
115.3208995,  115.3208995,   51.9965027,   51.9965027,   51.9965027,
51.9965027,   51.9965027,   51.9965027,  51.9965027,   51.9965027,
180.0496045, 180.0496045,  180.0496045,  180.0496045,  180.0496045,
180.0496045,  180.0496045,  180.0496045,  791.3223240,  791.3223240,
791.3223240,  791.3223240,  791.3223240,  791.3223240,  791.3223240,
791.3223240, 1280.0179973, 1280.0179973, 1280.0179973, 1280.0179973,
1280.0179973, 1280.0179973, 1280.0179973, 1280.0179973,  728.9582154,
728.9582154,  728.9582154,  728.9582154,  728.9582154,  728.9582154,
728.9582154,  728.9582154,  752.4591144,  752.4591144,  752.4591144,
752.4591144,  752.4591144,  752.4591144,  752.4591144,  752.4591144)

test <- data.frame(YEARS,MOISTURE,VARIANCE)

mod.nls <- nls(MOISTURE ~ A/(1+B*exp(-k*YEARS)),
	data = test, 
	weights = 1/VARIANCE,
	start = list(A=1500, B=200, k=0.03345),
	control=list(maxiter = 500),
	trace=TRUE)
summary(mod.nls)

Following the example of pdf!

st1 <- expand.grid(A = seq(0, 2000, len = 10),
B = seq(0, 500, len = 10), k = seq(-1, 10, len = 10))
st1

mod.nls2 <-nls2(MOISTURE ~ A/(1+B*exp(-k*YEARS)),
	data = test, 
	start = st1,
        algorithm="brute-force")
mod.nls2

I appreciate everyone's attention.
-- 
View this message in context:
http://r.789695.n4.nabble.com/How-using-the-weights-argument-in-nls2-tp2524328p2524328.html
Sent from the R help mailing list archive at Nabble.com.

On Thu, Sep 2, 2010 at 11:09 AM, Ivan Allaman <ivanalaman at yahoo.com.br>
wrote:
>
> Good morning gentlemen!
>
> How using a weighted model in nls2? Values with the nls are logical since
> values with nls2 are not. I believe that this discrepancy is due to I did
> not include the weights argument in nls2.
>
Just to follow up, nls2 was ignoring the weights argument.  This is
now fixed in the development version of nls2.  The weights and no
weights versions below give different residual sum of squares showing
that weights is not ignored:

library(nls2)
# grab development version
source("http://nls2.googlecode.com/svn/trunk/R/nls2.R")
BOD2 <- cbind(BOD, w = 1:6)
nls2(demand ~ a + b*Time, data = BOD2, start = c(a = 1, b = 1),
   weights = w, alg = "brute")

# compare against
nls2(demand ~ a + b*Time, data = BOD2, start = c(a = 1, b = 1),
   alg = "brute")

-- 
Statistics & Software Consulting
GKX Group, GKX Associates Inc.
tel: 1-877-GKX-GROUP
email: ggrothendieck at gmail.com