Exploratory Analysis

Exploratory analysis methods examine data for trends, correlations, or other relationships. Sometimes, models are created which can later identify when new data does not follow the same trend as previous data (see, for example, using principal components analysis in multivariate statistical process control, see: Process Control and Statistics) or can be used to predict an amount of material or property (which is also discussed in Quantitative Regression Analysis.) Often, however, these methods are used simply to learn more about the data.

Top-Level Exploratory Analysis Functions

These functions provide high-level analysis of data. Most have various options and output model structures.

analysis - Graphical user interface for data analysis.

pca - Principal components analysis.

mcr - Multivariate curve resolution with constraints.

purity - Self-modeling mixture analysis method based on purity of variables or spectra.

cluster - Cluster analysis with dendrograms using various algorithms.

corrspec - Correlation spectroscopy maps.

crossval - Cross-validation for decomposition and linear regression.

See Also Multiway Exploratory Analysis

Evolving and Windowed Factor Analysis

These function provide moving and "evolving" (growing) windowed analysis of data.

evolvfa - Evolving factor analysis (forward and reverse).

ewfa - Evolving window factor analysis.

wtfa - Window target factor analysis.

Other Exploratory Tools

These are data-exploration tools, some of which provide interfaces to analyze the data or other medium-level analysis functionality.

mlpca - Maximum likelihood principal components analysis.

coda_dw - Calculates values for the Durbin_Watson criterion of columns of data set.

coda_dw_interactive - Interactive version of CODA_DW.

comparelcms_sim_interactive - Interactive interface for COMPARELCMS.

trendtool - Univariate trend analysis tool.

estimatefactors - Estimate number of significant factors in multivariate data.

Application of Models to New Data

In most cases, the function used to create a model (e.g. PCA, PLS, etc) is also used to make a prediction with the created model. See the function used for more information on this. In addition, these utilities may be of use for certain applications.

modelselector - Create or apply a model selector model.

compressmodel - Remove references to unused variables from a model.

matchvars - Align variables of a dataset to allow prediction with a model.

pcapro - Projects new data on old principal components model.

Model Analysis and Calculation Utilities

Low-level engine and calculation functions.

qconcalc - Calculate Q residuals contributions for predictions on a model.

residuallimit - Estimates confidence limits for sum squared residuals.

reviewmodel - Examines a standard model structure for typical problems.

tconcalc - Calculate Hotellings T2 contributions for predictions on a model.

tsqlim - Confidence limits for Hotelling's T^2.

varcap - Variance captured for each variable in PCA model.

varimax - Orthogonal rotation of loadings.

als - Alternating Least Squares computational engine.

datahat - Calculates the model estimate and residuals of the data.

dispmat - Calculates the dispersion matrix of two spectral data sets.

pcaengine - Principal Components Analysis computational engine.

tsqmtx - Calculates matrix for T^2 contributions for PCA.

comparelcms_simengine - Calculational Engine for comparelcms.

Plotting Utilities

modlrder - Displays model info for standard model structures.

plotloads - Extract and display loadings information from a model structure.

plotscores - Extract and display score information from a model.

ploteigen - Builds dataset object of eigenvalues/RMSECV information.

ssqtable - Displays variance captured table for model.

(Sub topic of Qualitative_Exploratory_Analysis_and_Classification)