22 oct. 2015

Improving plots in the Monitor Function

I am trying to find the best way to check the performance of a model comparing the reference values with the predicted values and to see the efect of a bias adjustment, so after working on the function a plot is generated.
I will probably add two more plots, but I would not want to overcharge the plotted information.
I will see.
At the moment the info generated with the function is:

> monitor10c12(Muestra,HUM_NUT,HUM_ING,sortref=TRUE)
WARNING: More than 20 samples are needed to run the Validation 
Validation Samples  = 16 
RMSEP    : 0.742 
Bias     : -0.707 
SEP      : 0.233 
Corr     : 0.989 
RSQ      : 0.979 
Slope    : 0.907 
Intercept: 0.275 
RER      : 16.1   Fair 
RPD      : 6.17   Excellent 
BCL(+/-): 0.124 
      ***Bias adjustment is recommended***
Residual Std Dev is : 0.198 
    ***Slope adjustment is recommended***
Using  SEP as std dev the residual distibution is: 
  Residuals into 68%   prob (+/- 1SEP)    = 0     % = 0 
  Residuals into 95%   prob (+/- 2SEP)    = 3     % = 18.8 
  Residuals into 99.5% prob (+/- 3SEP)    = 7     % = 43.8 
  Residuals outside 99.5% prob (+/- 3SEP) = 9     % = 56.2 
  Samples outside UAL  = 0 
  Samples outside UWL  = 0 
  Samples inside   WL  = 3 
  Samples outside LWL  = 13 
  Samples outside LAL  = 9 
With Bias correction the Residual Distribution would be:
  Residuals into 68%   prob (+/- 1SEP)     = 13     % = 81.2 
  Residuals into 95%   prob (+/- 2SEP)     = 15     % = 93.8 
  Residuals into 99.5% prob (+/- 3SEP)     = 16     % = 100 
  Residuals outside  99.5% prob (> 3SEP)   = 0      % = 0 


No hay comentarios:

Publicar un comentario