Orthogonal Design 1. Introduction
Orthogonal experiment is to study the multiple factors and levels of one test method, which is designed to carry out the test, replace the whole by a small number of experimental use of orthogonal interview experience, according to the orthogonality of the orthogonal table from a comprehensive test pick the right, a representative test point, these points have a representative "homogeneously dispersed, comparable neat" characteristics. The factors discussed in this specification is the only independent orthogonal test, orthogonal test each interacting factors in our design of test use cases when less than, so I do not mention.
Orthogonal table is a special form, typically with L n (m K shown), L represents an orthogonal table, the number of rows and n represents the number of trials or orthogonal, k representative can arrange for the maximum number of factors impact indicators the number of columns or orthogonal, m represents the number of levels of each factor, and there are n = k * (m-1 ) +1. Orthogonal Test principle and characteristics, see Appendix .
2. Causes Orthogonal Test Method
For single factors or test two factors, because of its small factor test design, implementation and analysis are relatively simple. However, in practical work , often need test also investigated the factors above three or three, if the comprehensive test, the test of a large scale, due to time and cost we can not conduct a comprehensive test, but the specific pick one of which test cases to test our hearts in doubt, do not always worry about the selection of those test cases will miss some serious flaws. To be effective, a reasonable reduction of working hours and the cost of the test, we use orthogonal experiment method to design test cases. Orthogonal test is scheduled multi-factor test, looking for an efficient optimal level combination of experimental design methods.
We use the test case to illustrate the benefits of using orthogonal test design test cases.
Test requirements:
Communications Department of a university were two classes, just finished a course, want to "sex", "class" and "achievements" of the three conditions of the query results communication system this course distribution, gender or class ratio of personnel inquiries:
Search "male and female" in accordance with "gender" =
According to "class" = "Class 1, Class 2," Query
According to "score" = "pass, fail" Query
According to the traditional design - All tests
Analysis of the above test requirements, there are three test element, the test element we call factors, each has two values, which we call level value, so all the number of test cases is 2 * 2 * 2 = 8 , see table below
| No. |
gender |
class |
Achievement |
| 1 |
Female |
1 class |
Passing |
| 2 |
Female |
1 class |
failed |
| 3 |
Female |
Class 2 |
Passing |
| 4 |
Female |
Class 2 |
failed |
| 5 |
male |
1 class |
Passing |
| 6 |
male |
1 class |
failed |
| 7 |
male |
Class 2 |
Passing |
| 8 |
male |
Class 2 |
failed |
The use of orthogonal test design, the number of test cases we get n = 3 * (2-1) + 1 = 4, for a two level factors there are just three L . 4 (2 . 3 ) orthogonal table can be applied , then come with four orthogonal test test method is as follows:
| No. |
gender |
class |
Achievement |
| 1 |
Female |
1 class |
Passing |
| 2 |
Female |
Class 2 |
failed |
| 3 |
male |
1 class |
failed |
| 4 |
male |
Class 2 |
Passing |
The actual test cases need to be supplemented on the basis of use cases designed by the orthogonal test.
4 test cases as compared with the number of test eight test cases is reduced. The larger the number the number of factors and levels more able to reflect the benefits of using orthogonal table. For example: For a four-factor test and each factor are three levels, if carried out in accordance with the comprehensive test requires 3 * 3 * 3 * 3 = 81 times. However, if L selected by Orthogonal 9 (3 . 4 ) orthogonal table, n = 4 * (3-1) + 1 = 9 tests can cover. This point can be illustrated by Orthogonal efficiently, rationally and reduce the man-hour test, the test cost savings.
Advantage : According to the orthogonality of the whole interview selected test out some representative test point, these characteristics have a representative "homogeneously dispersed, comparable neat" characteristics. By reducing the use of orthogonal test test case, reasonable working hours and reduce the cost of testing and improve the effectiveness of the test case. It is a highly efficient, rapid and economical method of experimental design.
Drawback : equal treatment for each state point, unfocused, easily lead the user does not frequently used features or scenarios, spend a lot of time to test design and execution, and in the use of important but not the focus of the test path.
Although orthogonal design have these shortcomings, but it can find the optimal level through a combination of part of the test, and therefore very popular with the actual work favored persons.