Properly document the semantics of event weights

[Juergen Reuter]

Weighted events do not give the expected behaviour -- namely, being between 0 and 1 for the events files. This has been destroyed in r2187. Why???

[ohl]

As I see it, there are four choices (<...> denotes average)

1. <w_i> = 1 (i.e. <sum w_i> = N)
2. <w_i> = sigma (i.e. <sum w_i> = N*sigma)
3. <w_i> = 1/N (i.e. <sum w_i> = 1)
4. <w_i> = sigma/N (i.e. <sum w_i> = sigma)

neither of which guarantees that 0 <= w <= 1 (which is one of the motivations for weighted events, btw.). WK has implemented a choice between 1. and 2. using the variable normalize_weight. I would prefer 4. over 2., but this is trivial and only needs to be documented properly.

[Juergen Reuter]

I would also prefer the choice 4. over 2. And agree on a proper documentation. This is probably our biggest task for the next two weeks.

[kilian]

Weighted events are actually broken in W2: channel selection doesn't work. I'll fix this first.

[kilian]

OK, weighted events should be fixed now [2191].

[kilian]

r2192 implements all options that Thorsten has listed, so it's now the user's choice.

The default is $event_normalization = "auto" which normalizes unweighted events to unity, and weighted events to the cross section, i.e., choices 1 and 2.

IMHO, 1 and 2 are preferable over 3 and 4, because (1) they are required in the LHEF standard, and (2) the normalization is independent of the number of events, so event samples can be cut or expanded without normalization change.

In fact, when choosing 3 or 4, appending to an event file should be an error, but I haven't implemented this.

Comments?

[Juergen Reuter]

I will try and write up some Easter bunny up today such that WK can concentrate on his technical stuff.

[Juergen Reuter]

Most things done in r2262. There are just a few things in the SINDARIN chapter missing, and a few remarks about the options of simulate in the event generation chapter. Will do that today. Details about the different event formats should be in the manual, too, but for me are part of the general manual ticket.

[Juergen Reuter]

Documentation done in r2265.Closing.

[Juergen Reuter]

Milestone v2.0.0final deleted