model-report_ge_weight-method: model method for report_ge_weight' this method uses samples...
In stacomiR: Fish Migration Monitoring

model,report_ge_weight-method

R Documentation

model method for report_ge_weight' this method uses samples collected over the season to model the variation in weight of glass eel or yellow eels.

Description

model method for report_ge_weight' this method uses samples collected over the season to model the variation in weight of glass eel or yellow eels.

Usage

## S4 method for signature 'report_ge_weight'
model(object, model.type = "seasonal", silent = FALSE)

Arguments

`object`	An object of class report_ge_weight-class
`model.type`	default 'seasonal', 'seasonal1','seasonal2','manual'.
`silent`	Default FALSE, if TRUE the program should no display messages

Details

Depending on model.type several models are produced

model.type='seasonal'. The simplest model uses a seasonal variation, it is fitted with a sine wave curve allowing a cyclic variation w ~ a*cos(2*pi*(d'-T)/365)+b with a period T. The modified day d' used is this model is set at 1 the 1st of august doy = d' + d0; d0 = 212, doy=julian days
model.type='seasonal1'. A time component is introduced in the model, which allows for a long term variation along with the seasonal variation. This long term variation is is fitted with a gam, the time variable is set at zero at the beginning of the first day of observed values. The seasonal variation is modeled on the same modified julian time as model.type='seasonal' but here we use a cyclic cubic spline cc, which allows to return at the value of d0=0 at d=365. This model was considered as the best to model size variations by Diaz & Briand in prep. but using a large set of values over years.
model.type='seasonal2'.The seasonal trend in the previous model is now modelled with a sine curve similar to the sine curve used in seasonal. The formula for this is sin(omega vt) + cos(omega vt), where vt is the time index variable omega is a constant that describes how the index variable relates to the full period (here, 2pi/365=0.0172). The model is written as following w~cos(0.0172*doy)+sin(0.0172*doy)+s(time).
model.type='manual'. The dataset don (the raw data), coe (the coefficients already present in the database, and newcoe the dataset to make the predictions from, are written to the environment envir_stacomi. please see example for further description on how to fit your own model, build the table of coefficients, and write it to the database.