The Partial Least Squares (PLS) regression
[1,2,3] node in SciCraft has the
following node arguments:
- Input ports:
- X: The input data matrix (Covariate values). Each row is an object and each
column contains the different variables (features)
- Y: The input response vector or matrix (multi-response)
- Input arguments:
- NumberOfComponents: Integer with the number of components to include in the
model (see below)
- Validation: Text that specifies the Validation type used on model;
- none: No Cross-Validation is performed.
- Leave-one-out: Leave one out cross-validation performed
- Random segments:
-fold cross-validation performed.
is an
argument given in ValidationSegments (See below)
- ValidationSegments: Integer. Number of folds used in the
Cross-validation if Validation is set to Random segments.
- JackKnife: Text with the following choices
- none: No jackknife performed on fitted model.
-scores on loadings:
scores on coefficients:
- ForceComponents: Text with the two following choices
- Find optimum: The number of components used are found by
minimising the Cross-Validation error.
- Use NumberOfComponents: Use the specified number of components.
- FDR: Text with the following choices:
- Compute FDR
-values: Compute False Discovery Rate (FDR)
-values, requires
JackKnife set different from 'none'.
- none: No FDR
-values computed.
- ModifiedTtest: Integer specifying the input value of the
modified T-test. Must be in the range of
.
- permutaions: Integer specifying the number of permutations used when computing the
FDR
-values.
- Output ports:
- diagnostics: An R object containing the following
elements:
- optimumNumberOfComponents: Scalar with the optimal number of components
used in the model. Based on the minimum Cross-Validation error.
- qValue: Vector containing the estimated FDR
-values for the covariates .
- Tstat: Vector containing the estimated
-values.
- Ypred: Vector containing the predicted response values, based
on the input covariate matrix
.
- plsModel: An R object containing the following
elements:
- loadings (data): Matrix containing the loadings.
- method: String contain the method used
- model: The Covariate matrix
and response
vector
.
- ncomp: Scalar with the number of components used in
the model.
- projection (data): Matrix containing the the projections
of the data onto the terms of a linear model
- residuals (data): Matrix with residuals. The
'th
column contains the residual vector for the model fitted with
components.
- scores (data): Matrix containing the scores (see below)
- validation: An R object with the results
of the Cross-validation containing the
following elements:
- adj: Vector or matrix adjustment values for
calculating bias corrected Mean Squared Error of Prediction
(MSEP).
- Method: String with the text ``CV'', for Cross-Validation
- MSEP: Vector or matrix of MSEP values for models
with
components. If the output is a matrix,
each row corresponds to one response variable
- MSEP0: Integer or vector with MSEP values (one for each response variable) for
a model with zero components, i.e., only the intercept.
- pred: Matrix containing the predicted response
values. Column number
gives the response value for the model
fitted with
number of components.
- R2: Matrix of unadjusted
values for models with
components. Each row corresponds to one response variable.
- Segments: List of segments used in the Cross-Validation.
- Xmeans: Vector containing the mean values of each
row of the input covariate matrix
.
- Xtotvar: Scalar that states the total variance of the
input matrix
.
- Xvar: List of scalars that explains how much variance is
explained by each component.
- Yloadings: Vector with the loadings to each of the
elements in in the response vector
.
- Ymeans: Scalar with the mean value of the response input
vector
- Yscores: Matrix with scores of the response vector
where each column
is the score for the model
fitted with
components.
Bjørn Kåre Alsberg
2006-04-06