Compound and Group Filters

[Red Devil]

Stan,

I have been working with compound filters rather than group filters as you can use both "AND" & "OR" functions in a compound filter. However, I like the idea of being able to select "At least x to x sub-filters must pass. Before I spend too much time developing a group filter can you tell me if it is OK to include compund filters inside a group filter. I suspect that it is OK, I just don't want to waste time if it's not.

[Bobes]

[quote=Red Devil]
Stan,

I have been working with compound filters rather than group filters as you can use both "AND" & "OR" functions in a compound filter. However, I like the idea of being able to select "At least x to x sub-filters must pass. Before I spend too much time developing a group filter can you tell me if it is OK to include compund filters inside a group filter. I suspect that it is OK, I just don't want to waste time if it's not.
[/quote]
You can apply both type of filters togehter: It means that any compound filter can be embedded into group filter.
GENERALLY
Compound filters are suitable in particular in cases where at least two situations can occur. Example: Filters→WN Hist→ SUM
Let´s assume that margins are Min=328 and Max=478 for the SUM1. Then if the compound filter will contain the following two filters WNHist→SUM Min= 328 - MAX = 350 AND WNHist→SUM Min= 440 - MAX = 478
the resulting package should contain only those tickets whose sum1 is either in deep decrease (from 328 to 350) or in high increase (from 440 to 478) . However the logical condition must be set to AND in this case. Setting this logical operator to OR will remove all the ticket from the package (both situation are not allowed at the same time - the first situation will exclude the second one).

Group filter allows for certain doubts in our estimation due to its condition at least "xx" TO "yy" must pass. This usually result in normal distribution as for the estimated parameter(s). In other words using this filter we admit that some of our estimations may not apply, however we are not able to say which out of estimated settings are incorrect.

Naturally, when you put filters together you must take into account their nature. Some simple examples are : You could not expect EVEN value for SUM if you chose the ODD/EVEN ratio = 3/3 at the same time; you can not expect ticket with numbers beginning at "1" and ending with "49" if you chose small value for StdDev. I am sure you know many other more complex examples.

ONCE AGAIN . You can combine both type of filters and built group filters in compound filters and vice versa

The best way how to estimate behaviour for built in filters is to try them separately in advance.

The best way how to create complex compound filters is to draw their tree structures with filters in blocks on A4 sheet before you start with their construction.

[Red Devil]

Bobes,

Thanks