Minimizemodel and Vipnway: Difference between pages

From Eigenvector Research Documentation Wiki
(Difference between pages)
Jump to navigation Jump to search
imported>Donal
(Created page with "===Purpose=== Shrinks model by removing non-critical information. ===Synopsis=== :mm = minimizemodel(model) %compress model :minimizemodel(model) %display size informatio...")
 
imported>Scott
No edit summary
 
Line 1: Line 1:
===Purpose===
===Purpose===


Shrinks model by removing non-critical information.
Calculate Variable Importance in Projection from NPLS model.


===Synopsis===
===Synopsis===


:mm = minimizemodel(model) %compress model
:vip_scores = vipnway(model)
:minimizemodel(model)      %display size information only


===Description===
===Description===


Models contain both the information necessary to apply that model to new data and also the results calculated with the model was built (such as scores, cross-validation results, Hotellings T^2, sum squared residuals from the calibration samples.) Although this additional calibration sample information is necessary to review the model results, they are not necessary to apply the model to new data.
Variable Importance in Projection (VIP) scores estimate the importance of each variable in the projection used in a NPLS model and is often used for variable selection. A variable with a VIP Score close to or greater than 1 (one) can be considered important in given model. Variables with VIP scores significantly less than 1 (one) are less important and might be good candidates for exclusion from the model. It works for X n-way and Y up to two-way and it assume samples are in the first mode.
 
MINIMIZEMODEL attempts to compress a model by removing the fields which are not strictly necessary to apply the model. Such compression will prevent the direct comparison of new sample results to calibration sample results, but the model will still be functional for on-line use, for example.
 
The extent of compression varies greatly between model types and will generally be more effective on models built from large numbers of samples and fewer variables as compared to models built from large numbers of variables and fewer samples.
 
If no outputs are requested, the sizes of all model fields with more than 100 bytes in size are returned.


====Inputs====
====Inputs====


* '''model''' = standard model structure to compress.
* '''model''' = A NPLS model structure from a NPLS model.


====Outputs====
====Outputs====


* '''mm''' = minimized model.
* '''vip_scores''' = a cell array with dimensions of: [modes 2 to n X # of columns in Y]. The first row in the cell array corresponds to VIP Scores for mode 2. The second row corresponds to VIP Scores for mode 3.
 


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


[[compressmodel]], [[modelstruct]]
[[selectvars]], [[genalg]], [[ipls]], [[plotloads]], [[pls]], [[plsda]], [[sratio]], [[rpls]], [[vip]]

Revision as of 17:06, 18 December 2018

Purpose

Calculate Variable Importance in Projection from NPLS model.

Synopsis

vip_scores = vipnway(model)

Description

Variable Importance in Projection (VIP) scores estimate the importance of each variable in the projection used in a NPLS model and is often used for variable selection. A variable with a VIP Score close to or greater than 1 (one) can be considered important in given model. Variables with VIP scores significantly less than 1 (one) are less important and might be good candidates for exclusion from the model. It works for X n-way and Y up to two-way and it assume samples are in the first mode.

Inputs

  • model = A NPLS model structure from a NPLS model.

Outputs

  • vip_scores = a cell array with dimensions of: [modes 2 to n X # of columns in Y]. The first row in the cell array corresponds to VIP Scores for mode 2. The second row corresponds to VIP Scores for mode 3.

See Also

selectvars, genalg, ipls, plotloads, pls, plsda, sratio, rpls, vip

Retrieved from "https://www.wiki.eigenvector.com/index.php?title=Minimizemodel&oldid=10233"