predict.lmcal, predict.nlscal {quantchem} | R Documentation |
Inverse predict concentration from responses, using all fitted calibration models.
predict.lmcal(object, dataset, conf.int = 0.95, ...) predict.nlscal(object, dataset, ...)
object |
an object of class 'lmcal' or 'nlscal', respectively |
dataset |
a vector of responses |
conf.int |
confidence intercal (only for lmcal) |
... |
additional arguments, currently ignored |
For linear models, the concentrations are calculated by inverse.predict()
, which
calls polyroot()
on modified polynomial coefficients. For nonlinear models,
concentrations are calculated with appropriate 'inverse' formulas.
A list containing following elements. Each element is a list of concentration vectors, calculated from a model, with name referring to the model.
fitted |
Concentrations calculated by fitted model |
upper |
Upper limit of confidence interval of inverse prediction |
lower |
Lower limit of confidence interval of inverse prediction |
The confidence interval for prediction is calculated by taking standard error of
prediction and dividing it by slope of calibration curve (estimated by derivative
)
Then, proper confidence interval is constructed using t statistic.
Lukasz Komsta
set.seed(1234) x=rep(1:10,10) y=jitter(sqrt(x)) fit=lmcal(x,y) predict(fit,rnorm(10,mean=2,sd=0.1))