chemosensors 0.7.6

Class SensorModel.

Description

Class SensorModel predicts a sensor signal in response to an input concentration matrix by means of a regression model stored in slot dataModel.

Details

The model explicitely assumes that the sensor response to a mixture of analytes is a sum of responses to the individual analyte components. Linear models mvr and plsr follow this assumtion in their nature.

Slots of the class:

num Sensor number (1:17). The default value is 1.
gases Gas indices.
ngases The number of gases.
gnames Names of gases.
concUnits Concentration units external to the model, values given in an input concentration matrix.
concUnitsInt Concentration units internal for the model, values used numerically to build regression models.
dataModel Data model of class SensorDataModel performs a regression (free of the routine on units convertion, etc).
coeffNonneg Logical whether model coefficients must be non-negative. By default, FALSE.
coeffNonnegTransform Name of transformation to convert negative model coefficients to non-negative values.
beta (parameter of sensor diversity) A scaling coefficient of how different coefficients of SensorDataModel will be in comparision with those coefficients of the UNIMAN sensors. The default value is 2.

Methods of the class:

predict Predicts a sensor model response to an input concentration matrix.
coef Extracts the coefficients of a regression model stored in slot dataModel.

The plot method has two types (parameter y):

response (default) Shows the sensitivity curves per gas in normalized concentration units.
predict Depicts input (parameter conc) and ouput of the model for a specified gas (parameter gases).

Examples


# sensor model: default initialization
sm <- SensorModel()

# get information about the model
show(sm)



 Sensor Model (num 1), beta 2, data model 'ispline'



print(sm)



 Sensor Model
 - num 1 
 - beta 2 
 - 3 gases A, B, C 
 - (first) data model
 - method: ispline (type: spline)
 - sensor model: coeffNonneg TRUE
   -- coefficients (first): 7.5816, 2.5812, 0 ... 0 




print(coef(sm)) # sensitivity coefficients



          [,1]
 [1,] 7.581598
 [2,] 2.581216
 [3,] 0.000000
 [4,] 7.194504
 [5,] 0.946807
 [6,] 0.000000
 [7,] 7.869399
 [8,] 1.551513
 [9,] 0.000000




plot(sm)  






# get available model names
model.names <- SensorModelNames()
print(model.names)



[1] "plsr"         "mvr"          "broken-stick" "ispline"     




# sensor model: custom parameters
sm <- SensorModel(num=7, model="plsr", gases=c(1, 3))

print(sm)



 Sensor Model
 - num 7 
 - beta 2 
 - 2 gases A, C 
 - (first) data model
 - method: plsr (type: mvr)
 - ncomp: 2 
 - sensor model: coeffNonneg FALSE
   -- coefficients (first): 0.186, 0.0116 




#plot(sm, uniman=TRUE) # add UNIMAN reference data (the model was build from)

# method plot
#  - plot types 'y': response, predict
sm <- SensorModel() # default sensor model

plot(sm, "response", main="plot(sm, 'response')") # default plot type, i.e. 'plot(sm)' does the same plotting






conc <- concSample(sm, "range", gases=1, n=10)
plot(sm, "predict", conc, gases=1, main="plot(sm, 'predict', conc, gases=1)")

See also

UNIMANshort