Find definitions and interpretation guidance for every design optimality statistic.

The condition number measures the collinearity among model terms. When you compare designs, a smaller condition number is better.

Use the condition number to compare different optimal designs or to compare the same design with different terms. A condition number of 1 indicates that the model terms are orthogonal. Larger values indicate more collinearity.

Most optimal designs are not orthogonal. Because terms in the model are not independent, the interpretation of non-orthogonal designs is less straightforward than the interpretation of orthogonal designs.

In these results, the condition number indicates that the data exhibit moderate to strong collinearity.

Factorial design selected according to D-optimality
Number of candidate design points: 64
Number of design points in optimal design: 32
Model terms: A, B, C, D, AB, AC, AD, BC, BD, CD
50% of the points in initial design are generated randomly
Remaining points added to initial design by Sequential method
Initial design improved by Exchange method
Number of design points exchanged is 1

Optimal Design
Row number of selected design points: 46, 54, 36, 8, 44, 47, 31, 55, 30, 43,
38, 59, 62, 15, 56, 24, 42, 20, 32, 16, 6, 45, 19, 17, 25, 49, 64, 10, 37,
1, 39, 3
Condition number: 259.114
D-optimality (determinant of XTX): 7.92282E+28
A-optimality (trace of inv(XTX)): 12.1719
G-optimality (avg leverage/max leverage): 0.96875
V-optimality (average leverage): 0.96875
Maximum leverage: 1

D-optimality indicates the design's ability to obtain precise estimates or predictions. When you compare designs, a larger D-optimality value is better.

You can use optimality metrics to compare designs, but remember that the optimality of a given D-optimal design is model dependent. That is, optimality is defined for a fixed design size and for a particular model.

In these results, the first optimal design has 25 design points and the second optimal design has 20 design points. The first design has a higher D-optimality statistic than the second optimal design, which is expected with more runs.

Response surface design augmented according to D-optimality
Number of candidate design points: 30
Number of design points to augment/improve: 20
Number of design points in optimal design: 25
Model terms: Block, A, B, C, D, AA, BB, CC, DD, AB, AC, AD, BC, BD, CD
Initial design augmented by Sequential method
Initial design improved by Exchange method
Number of design points exchanged is 1
Optimal Design
Row number of selected design points: 1, 3, 4, 6, 8, 9, 10, 13, 15, 16, 17, 19,
22, 23, 24, 25, 26, 27, 28, 30, 2, 5, 14, 18, 20
Condition number: 8.53018
D-optimality (determinant of XTX): 3.73547E+20
A-optimality (trace of inv(XTX)): 1.99479
G-optimality (avg leverage/max leverage): 0.64
V-optimality (average leverage): 0.64
Maximum leverage: 1

Response surface design selected according to D-optimality
Number of candidate design points: 30
Number of design points in optimal design: 20
Model terms: Block, A, B, C, D, AA, BB, CC, DD, AB, AC, AD, BC, BD, CD
Initial design generated by Sequential method
Initial design improved by Exchange method
Number of design points exchanged is 1
Optimal Design
Row number of selected design points: 22, 23, 25, 27, 4, 8, 19, 2, 14, 15, 13,
6, 9, 3, 16, 24, 28, 30, 26, 1
Condition number: 10.2292
D-optimality (determinant of XTX): 2.73819E+18
A-optimality (trace of inv(XTX)): 2.50391
G-optimality (avg leverage/max leverage): 0.8
V-optimality (average leverage): 0.8
Maximum leverage: 1

A-optimality measures the average variance in the regression coefficients of the fitted model. When you compare designs, a smaller A-optimality value is better.

You can use optimality metrics to compare designs, but remember that the optimality of a given A-optimal design is model dependent. That is, optimality is defined for a fixed design size and for a particular model. Designs that are more D-optimal are not necessarily more A-optimal.

In these results, the first optimal design has 25 design points and the second optimal design has 20 design points. The first design has a lower A-optimality statistic than the second optimal design, which is expected with more runs.

G-optimality is the ratio of the average prediction variance to the maximum prediction variance over the design points. G-optimal designs minimize the denominator, while V-optimal designs minimize the numerator. Ideally, you want both the numerator and denominator to be smaller values.

You can use optimality metrics to compare designs, but remember that the optimality of a given G-optimal design is model dependent. That is, optimality is defined for a fixed design size and for a particular model. Designs that are more D-optimal are not necessarily more G-optimal.

In these results, the first optimal design has 25 design points and the second optimal design has 20 design points. The design with more points is less G-optimal than the design with more points, even though the larger design is more D-optimal.

V-optimality measures the average prediction variance over the set of design points. When you compare designs, a smaller V-optimality value is better.

You can use optimality metrics to compare designs, but remember that the optimality of a given V-optimal design is model dependent. That is, optimality is defined for a fixed design size and for a particular model. Designs that are more D-optimal are not necessarily more V-optimal.

In these results, the first optimal design has 25 design points and the second optimal design has 20 design points. The first design has a lower V-optimality statistic than the second optimal design, which is expected with more runs.

Maximum leverage indicates that a design has a highly influential point when the maximum leverage is much larger than V-optimality. Minitab uses this value in the denominator when calculating G-optimality.

Use maximum leverage to determine when a design contains at least one influential point. Designs that are more D-optimal can have influential points.

In these results, the maximum leverage is 1 and the V-optimality is 0.8. In this optimal design, none of the factor levels in row 2 are in any of the other points.

Minitab displays the largest and smallest distances between the selected design points. This value is the Euclidean distance.

The difference between the largest and the smallest distance values indicates how uniformly the points are spread in the design space. You can use this information to compare designs.