Partial Least Squares - Prediction
main topic      how to     see also 

Stat > Regression > Partial Least Squares > Prediction

You can calculate and store predicted response values using your PLS model for two purposes: testing prediction quality and predicting new responses.

·    Testing prediction quality: You can apply your PLS model to a new data set that includes responses for each observation. Minitab calculates new response values and compares them to the actual response for each observation, calculating a test R2, which indicates the model's predictive ability.

·    Predicting new responses: You use the PLS model to calculate new response values for a set of predictors for which you have no response data. Without response data, a test R2 cannot be calculated.

Dialog box items

New observation for continuous predictors: Enter the new observation for each continuous predictor in the same order that each predictor is entered in the model. You can enter one numeric value for each predictor or one numeric column of new observations for each predictor. Columns must be the same length.

New observation for categorical predictors: Enter the new observation for each categorical predictor in the same order that each predictor is entered in the model. You can enter one value for each predictor or one column of new observations for each predictor. Columns must be the same length. If you type a new observation, you must enclose text values in double quotes (e.g., "Female").

New observation for responses (Optional): Enter the numeric columns containing the response values. You cannot type the response values; they must be stored in columns. The number of response columns must equal the number of responses in the model and be the same length as the predictors containing new observations.

Confidence level: Type the desired confidence level (for example, type 90 for 90%). The default is 95%.

Storage

Fits: Check to store fitted values for new observations.

SE of fits: Check to store estimated standard errors of the fitted values.

Confidence limits: Check to store lower and upper limits of the confidence interval for the prediction.

Prediction limits: Check to store lower and upper limits of the prediction interval.