TMPCA Results - Quick Tab

Select the Quick tab of the TMPCA Results dialog box to access the options described here.

Note: For specific details on R2X and Q2, Hotelling T2, and other technical functions mentioned in the option descriptions below, see PCA and PLS Technical Notes.
Element Name Description
Summary Click this button to generate a summary spreadsheet for the principal components. This output includes the R2X and Q2 statistics, the vector of eigenvalues, number of iterations, and other forms of useful statistics for model interpretation and review.
Summary overview Click this button to generate a histogram of cumulative R2X and Q2.
Variable importance Click this button to produce a spreadsheet of variable importance. The measure of importance of a variable is given by its modeling power. A variable with a modeling power equal to one is completely relevant for building the TMPCA model. Variables with modeling power close to "number of components" divided by "number of variables" are regarded to be less significant.
Variable importance Click this button to create a histogram of variable importance.
Sort variables by importance Select this check box to sort variables by their importance in the Variable importance spreadsheet and histogram (see above).
Control charts The following options are available in the Control charts group box:
T2 chart Click this button to create a Hotelling T2 chart for the individual batches in the data set together with the line plot of the control limit. Note that weak outliers may not show in T2. A better way of detecting weak outliers is distance-to-model. For TMPCA, each principal component has I t-scores (I is the number of batches). These are simply the projection of the original data in the direction of a principal component. In other words, it is a column of the T matrix. The t-scores are concerned with assessing the over all quality of a batch.
Click this button to produce a spreadsheet of the Hotelling T2 for the individual observations in the data set. Data cases with values of Hotelling T2 exceeding the control limit will be displayed in red. See the previous option for more details.
SPE (Q) chart Click this button to create an SPE (Q) (squares prediction error SPE) chart together with the line plot of the control limit. SPE represents noise (unstructured variation) in the data that cannot be captured by the model (since it does not follow a particular pattern). SPE (Q) error can be used for tracking the evolution of a batch in time. Thus, by observing the trend in SPE (Q) of a particular batch, we can monitor the evolution of a batch.
Click this button to produce an SPE (Q) (squares prediction error SPE) spreadsheet. Data cases with values of SPE exceeding the control limit will be displayed in red. See the previous option for more details.
Control limit Enter in this box the confidence limit for defining the upper and lower bounds of the control charts.
Warning limit Select this check box and enter in the adjacent box the warning limit.
Auto-fit more components by cross-validation You can use this option after adding or deleting one or more component from the TMPCA model. This action, which is equivalent to clicking the OK button in the TMPCA dialog box, invokes the v-fold cross-validation method to rebuild your model after you have manually added or removed one or more component. This option is available only if cross-validation was used to fit the model.