Re: Stochastic Queries
From: Joe Thurbon <usenet_at_thurbon.com>
Date: Mon, 10 Sep 2007 21:01:05 GMT
Message-ID: <lgiFi.33701$4A1.11655_at_news-server.bigpond.net.au>
>
> You can write some proprietary kludges, but frankly they have problems
> with skewed distributions. Better to stick to a stats package to draw
> your samples (with or without replacements? Tiered? etc.?) There was
> a good book years ago with a title somethign like a SAMPLER OF
> SAMPLING that went into the details.
>
> One trick I have used is to build a table:
>
> CREATE TABLE Samples
> (seq_nbr INTEGER NOT NULL PRIMARY KEY,
> random_nbr INTEGER NOT NULL); -- or whatever data type you need
>
> then load it with the particular random numbers I wanrted from a stat
> package. More work, but MUCH better results.
>
Date: Mon, 10 Sep 2007 21:01:05 GMT
Message-ID: <lgiFi.33701$4A1.11655_at_news-server.bigpond.net.au>
-CELKO- wrote:
>>> As a part of some experimental work I've been doing, I need to write what I think is best described as a stochastic quota query. <<
>
> You can write some proprietary kludges, but frankly they have problems
> with skewed distributions. Better to stick to a stats package to draw
> your samples (with or without replacements? Tiered? etc.?) There was
> a good book years ago with a title somethign like a SAMPLER OF
> SAMPLING that went into the details.
>
> One trick I have used is to build a table:
>
> CREATE TABLE Samples
> (seq_nbr INTEGER NOT NULL PRIMARY KEY,
> random_nbr INTEGER NOT NULL); -- or whatever data type you need
>
> then load it with the particular random numbers I wanrted from a stat
> package. More work, but MUCH better results.
>
Thanks -CELKO-,
I'll have a think about it.
Cheers,
Joe
Received on Mon Sep 10 2007 - 23:01:05 CEST