Specify the analysis options for Response Optimizer

Fit a model from the Predictive Analytics Module. Select Response Optimizer in the results. Select Setup.

Starting values and confidence intervals are available for binary logistic regression models and linear regresion models. Consideration of missing values is available for TreeNet^®, Random Forests^®, and MARS^® models.

In This Topic

Consider missing values during optimization
Constraints
Starting Values
Confidence level for all intervals
Type of confidence level

Consider missing values during optimization

Select this option so that the search for the optimal solution considers missing values. The search includes missing values for predictors that had missing values in the training data set during the construction of the model. Consider this option when missing values are meaningful for your application. Usually, missing values are not meaningful but can be in certain applications. For example, if missing values represent values below a detectable threshold for a continuous variable, then one interpretation of a missing value in the solution is to minimize that predictor in the application.

If you select this option then you can select Hold at missing as a constraint for predictors that have missing values in the training data.

Constraints

You can hold continuous and categorical variables at a specific value or limit the range of possible values.

Variable: Displays all the variables that are included in a model. This column does not take any input.
Constraint: For each variable, select No constraints, Hold at value, or Constrain to region. If the analysis allows missing values and the predictor had missing values in the training data, then the Hold at missing option is available.
Hold Value: For each variable that you specified Hold at value, enter a value at which to hold the variable. Minitab uses this value for the variable setting to calculate the fitted values.
Lower: For each variable that you specified Constrain to region, enter a minimum value. Minitab selects a value that is greater than or equal to this value.
Upper: For each variable that you specified Constrain to region, enter a maximum value. Minitab selects a value that is less than or equal to this value.

Starting Values

If the algorithm produces unacceptable results, you can try to improve the results by entering a starting point for the search algorithm. This option is available for binary logistic regression models and linear regression models.

Variable: Displays all the continuous variables that are included in a fitted model. This column does not take any input.
Starting Value: Enter a value for each continuous variable. Each value must be between the minimum and maximum observed values for that variable. If you enter a constraint for a variable, the starting value must satisfy that constraint. You cannot enter a starting value when you specify a hold value for a variable.

Confidence level for all intervals

Enter the level of confidence for the confidence intervals for the coefficients and the fitted values. Confidence levels are available for binary logistic regression and linear regression models.

Usually, a confidence level of 95% works well. A 95% confidence level indicates that, if you took 100 random samples from the population, the confidence intervals for approximately 95 of the samples would contain the mean response. For a given set of data, a lower confidence level produces a narrower interval, and a higher confidence level produces a wider interval.

Type of confidence level

You can select a two-sided interval or a one-sided bound. For the same confidence level, a one-sided bound is closer to the point estimate than the bounds of a two-sided interval. The upper bound does not provide a likely lower value. If you request an upper bound, then there is no lower bound. If you request a lower bound, then there is no upper bound.

For example, the predicted mean concentration of dissolved solids in water is 13.2 mg/L. The 95% confidence interval for the mean of multiple future observations is 12.8 mg/L to 13.6 mg/L. The 95% upper bound for the mean of multiple future observations is 13.5 mg/L, which is more precise because the bound is closer to the predicted mean.

Two-sided

Use a two-sided confidence interval to estimate both likely lower and upper values for the mean response.
Use a two-sided prediction interval to estimate both likely lower and upper values for a single future observation.

Lower bound

Use a lower confidence bound to estimate a likely lower value for the mean response.
Use a lower prediction bound to estimate a likely lower value for a single future observation.

Upper bound

Use an upper confidence bound to estimate a likely higher value for the mean response.
Use an upper prediction bound to estimate a likely higher value for a single future observation.