Ils esterror: Difference between revisions

From Eigenvector Research Documentation Wiki
Jump to navigation Jump to search
imported>Scott
No edit summary
imported>Jeremy
No edit summary
Line 5: Line 5:
===Synopsis===
===Synopsis===


:model = ils_esterror(model);
:[este,info] = ils_esterror(model);       %calibration estimation error
:pred  = ils_esterror(model,pred);
:[este,info] = ils_esterror(model,pred); %prediction estimation error


====Inputs====
====Inputs====
* '''model''' = a standard model structure (see PLS, PCR, MLR).
* '''model''' = a standard model structure (e.g. PLS, PCR, MLR).


====Optional Inputs====
====Optional Inputs====
Line 15: Line 15:


====Outputs====
====Outputs====
* model = ils_esterror(model);
* '''este'''  = estimation error for y. If (pred) was supplied, the estimation error for the prediction or validation is returned. Otherwise, the estimation error for the calibration samples is returned.
:The input (model) is output, with the field model.detail.esterror.dy populated with the estimation error.
* '''info'''  = string describing how the estimation was calculated (using RMSEC or RMSECV).
* pred = ils_esterror(model,pred);
 
:The input (pred) is output, with the field pred.detail.esterror.dy populated with the estimation error.
===Algorithm===
 
The estimate uses Eqn 9 of Faber, N.M. and Bro, R., Chemomem. and Intell. Syst., 61, 133-149 (2002). If available, MSECV is used instead of MSE [see output (info)]. Note from the ref.: "Faber [N.M. Faber, Chemom. Intell. Lab. Syst. 52 (2000) 123] has found the performance of Eq. (8)" (which is used to derive Eqn 9) "to rely heavily on the ability to correctly estimate the optimum model dimensionality.
 
Note that the estimate can be <0, in those cases it is set ==0.


===See Also===
===See Also===


[[analysis]], [[pcr]], [[pls]]
[[analysis]], [[pcr]], [[pls]]

Revision as of 11:17, 25 June 2010

Purpose

Estimation error for ILS models.

Synopsis

[este,info] = ils_esterror(model); %calibration estimation error
[este,info] = ils_esterror(model,pred); %prediction estimation error

Inputs

  • model = a standard model structure (e.g. PLS, PCR, MLR).

Optional Inputs

  • pred = a standard model prediction or validation structure [e.g. outputs (pred) or (valid) from PLS, PCR, or MLR].

Outputs

  • este = estimation error for y. If (pred) was supplied, the estimation error for the prediction or validation is returned. Otherwise, the estimation error for the calibration samples is returned.
  • info = string describing how the estimation was calculated (using RMSEC or RMSECV).

Algorithm

The estimate uses Eqn 9 of Faber, N.M. and Bro, R., Chemomem. and Intell. Syst., 61, 133-149 (2002). If available, MSECV is used instead of MSE [see output (info)]. Note from the ref.: "Faber [N.M. Faber, Chemom. Intell. Lab. Syst. 52 (2000) 123] has found the performance of Eq. (8)" (which is used to derive Eqn 9) "to rely heavily on the ability to correctly estimate the optimum model dimensionality.

Note that the estimate can be <0, in those cases it is set ==0.

See Also

analysis, pcr, pls