Weighting function that can be supplied to the weights argument of nlsLM or nls
wfct(expr)
expr |
An expression specifying the weighting scheme as described in the Details section below. |
The weighting function can take 5 different variable definitions and combinations thereof:
the name of the predictor (independent) variable
the name of the response (dependent) variable
error: if replicates y_{ij} exist, the error sigma_{ij}
fitted: the fitted values hat{y}_i of the model
resid: the residuals y_i - hat{y}_i of the model
For the last two, the model is fit unweighted, fitted values and residuals are extracted and the model is refit by the defined weights.
The results of evaluation of expr in a new
environment, yielding the vector of weights to be applied.
Andrej-Nikolai Spiess
### Examples from 'nls' doc ###
## note that 'nlsLM' below may be replaced with calls to 'nls'
Treated <- Puromycin[Puromycin$state == "treated", ]
## Weighting by inverse of response 1/y_i:
nlsLM(rate ~ Vm * conc/(K + conc), data = Treated,
start = c(Vm = 200, K = 0.05), weights = wfct(1/rate))
## Weighting by square root of predictor \sqrt{x_i}:
nlsLM(rate ~ Vm * conc/(K + conc), data = Treated,
start = c(Vm = 200, K = 0.05), weights = wfct(sqrt(conc)))
## Weighting by inverse square of fitted values 1/\hat{y_i}^2:
nlsLM(rate ~ Vm * conc/(K + conc), data = Treated,
start = c(Vm = 200, K = 0.05), weights = wfct(1/fitted^2))
## Weighting by inverse variance 1/\sigma{y_i}^2:
nlsLM(rate ~ Vm * conc/(K + conc), data = Treated,
start = c(Vm = 200, K = 0.05), weights = wfct(1/error^2))Please choose more modern alternatives, such as Google Chrome or Mozilla Firefox.