2 dic. 2015

Validation Report

Finally I was able to solve some issues with the data frames used in the monitor function, which converted the characters columns to factors so the sample identification was changed in case it contains not just numbers (as.is=TRUE will solve this problem). It was important also to extend the data frames correctly, this way the new data frame has the same structure all the time and everything worked fine.
The final report can be improved and new update could be added in the future, but at the moment quite happy with the results.
This could be a report for a protein validation in some kind of sausages measured by a NIT instrument and the reference method:

> monitor10c22dev(sausage,sortref=FALSE,eqalaxis=TRUE)
Validation Samples  = 88 
Reference Mean  = 22 
Predicted Mean  = 20.6 
RMSEP    : 2.49        (actual error)
Bias     : 1.39 
SEP      : 2.07        (error if we correct the Bias)
Corr     : 0.915 
RSQ      : 0.838 
Slope    : 0.884 
Intercept: 3.77 
RER      : 11.8   Poor 
RPD      : 2.46   Fair 
Residual Std Dev is : 2 (Error if we correct slope/intercept)
    ***Slope/Intercept adjustment is recommended***
BCL(+/-): 0.439 
    ***Bias adjustment in not necessary***
Without any adjustment and using  SEP as std dev the residual distibution is: 
  Residuals into 68%   prob (+/- 1SEP)    = 54     % = 61.4 
  Residuals into 95%   prob (+/- 2SEP)    = 83     % = 94.3 
  Residuals into 99.5% prob (+/- 3SEP)    = 87     % = 98.9 
  Residuals outside 99.5% prob (+/- 3SEP) = 1      % = 1.14 
With S/I correction and using Sres as standard deviation, the Residual Distribution would be:
  Residuals into 68%   prob (+/- 1Sres)     = 67     % = 76.1 
  Residuals into 95%   prob (+/- 2Sres)     = 84     % = 95.5 
  Residuals into 99.5% prob (+/- 3Sres)     = 87     % = 98.9 
  Residuals outside  99.5% prob (> 3Sres)   = 1      % = 1.14 
 
 

No hay comentarios:

Publicar un comentario