1. ## OATS

I have read several articles on orthogonal array testing. But I am not clear of one thing. When we select the optimized array, based on what criteria the levels are arrived at?
Consider the following example:
3 variables(a,b,c) which can have 3 values(1,2,3) each. There could be 27 possible test scenarios (3x3). Using OA, number of test cases are optimized to 9. What is the criteria for selecting these 9 cases? I could not find the selection criteria for these 9 cases anywhere.
Are any one aware of the details?

2. ## Re: OATS

Any takers for this question? Appreciate if any one who has understood OA well can give a good explanation.

3. ## Re: OATS

There have been some other threads on OA in the Functional Testing forum, have you done a check through those?

I've never utilized it as a tool, so I can't really add much to this.

- M

4. ## Re: OATS

michaeljf, Thanks for the response.

I have seen those threads &amp; did some extensive search on internet. But, my question remain unanswered. This may look simple to OA experts, but this basic question is bothering me. If any one has the answer, can they please respond?

Should I post this in 'Functional Testing' or any other forums? Please suggest.

5. ## Re: OATS

You could pose the question again, nothing wrong with that. Thinking about it outside of Orthogonal Testing though anything you choose I would assume would need to be the high risk scenarios if you can only choose a subset. That's basic to any listing of scenarios and where you have limited time or resources to check only some of them. How that is done within OA though, I don't know.

- M

6. ## Re: OATS

I do not understand your question:
For your 3x3 problem there exists an OA with 9 test cases which generates all pair combinations.
For these 9 test cases the levels for the 3 variables are specified in the corresponding OA.
So the criteria (for OATS) to get all pair combinations is met...

So what do you need in addition? Do you want to know the background behind the OA generation?
Do you want to know if there are more possible solutions for a specific OA-problem?
Etc.?

regards,
Thomas

7. ## Re: OATS

Hi,

First I will give you the solution, but then I will ask you to forget about orthogonal arrays in software testing.

Based on your example, you are stating that you have a function with 3 interdependent parameters, and that each parameter has 3 possible variable values. The Cartesian product is all possible combinations which is equal to 27.

Next you must go to a Taguchi array selector available at http://www.freequality.org (when their site is up...it seems to be mostly down when you really need it). In this example the number of interdependent parameters is 3 so P = 3. The number of variables per parameter is 3 so L = 3. Next you select P=3,L=3 in the chart and you will get the apropriate L(sub #) array. Finally, you manually map in each variable into the array.

Orthogonal arrays are one of those manufacturing things that some perverted mind tried to shoehorn into the software testing world. They are very labor intensive. Basically, they are complex solutions to complex problems. Orthogonal arrays also only work well when the number of variables for all parameters is equal (such as in your example). If the number of variables is different for each parameter then you end up trying to combine 2 or more arrays by overlaying them. This is truly a nightmare.

So, now please forget everything you know about orthogonal arrays with regards to software testing (other than perhaps they are mostly impractical) and take a look at combinatorial analysis tools such as PICT on http://www.pairwise.org.

#### Posting Permissions

• You may not post new threads
• You may not post replies
• You may not post attachments
• You may not edit your posts
•

vBulletin Optimisation provided by vB Optimise v2.6.0 Beta 4 (Pro) - vBulletin Mods & Addons Copyright © 2016 DragonByte Technologies Ltd.