Using orthogonal matrix to design test spec--2

  Today,I will go on talking the detail for it.First and foremost,I think I should illustrate that in what circumstances can we use orthogonal array testing strategy to design test spec.Except for confirming factors and levels,we also should make sure the interaction between each factor.In other words,they should have some conditionality.We can use strength to describe this relationship.Take an example,if there are three factor--A,B,C.If we consider the strenth is two,we should think the relationship between A and B,A and C,B and C respectively.In the same principle,we think the relationship among A,B,and C,as strenth is 3.So if there is no relationship between factors and there are only two factors,I think use this method is no meaning.After settle this issue,we can begin to discussing how to design test spec using this method.

  At first,we should know how many factors are they have.It determine how many column the table will have.Second,we should understand how many levels of each factor.There have a conception we should know.How to make sure levels?What is norm?The levels is not how many values we can input in fact.(This concept has something wrong in my forer article)It should determine base on the principle of divisiory equivalent.For instance,if we have a textbox which can input 11 characters.The levles we can find like these,12 chracters,11 characters or 0 characters.In other words,we won't input several reasonable values to test if it is correct.We divide the value to two scopes--reasonable and unresonable.In reaonable scope,we can input any correct value to test.Meanwhile,we can input only one incorrect value to test if it can judge.That's the reason you see in some resourse,they only test two instance--fill a value or not.After that,we can start our third step--choose orthogonal table--base on our factors and levels.The table should consistent with this two values.Then we can find some tables online,certainly,if levels is special,you can design it by yourself.For example,maybe one factor have two levles,other two factors have three levels or more.In this case,we need design it by ourself.For now,it can said we have finished a basically table.Furthermore,sometime,if a factor has odd levels,we can take this level of factor alternate in the table from the top to bottom. At last,determine how many runs it has?This is base on your mind.We also can add some rows to cover some uncertain instances.These are all up on different circs.

  Ok,here,we can said a comletely orthogonal table is finished.But,I believe the core of it need farther understanding.I won't take an instance in here,however,I can put a link,it will more helpful.And I also thanks the author who give me lots of suggestions on this question.His blog is very good.Thanks a lot!The link is http://blog.csdn.net/zeeslo/archive/2006/09/26/1289991.aspx