On Sun, 20 Jul 2003 11:56:00 +0100, "Shaun Williams"

> Hi everyone,

> I am very interested to use SPSS for a number of direct marketing type

> analysis. I have brought a couple of textbooks that talk through stats

> and/or spss but none appear to use examples drawn from direct marketing.

> Does anyone know of any books/articales which specifically describe direct

> marketing applications of spss?

> One of the things i would like to do is calculate the sample size for a

> number of mailing cells, the data would look like this;

> Marketing code (say 5123)

> Expected response rate (say 5.5%)

> Allowable margin of error (say 0.5%)

> These are column headinsg (i.e. variables) and I would have dozons rows, one

> for each marketing code. I know there a way of calculating the required

> sample size using these two varaibles, but I am not sure how to do it in

> spss.

I think the SPSS readers know SPSS and stats, but very little

marketing -- so none of us recognize your question.

Even if SPSS has something built in for that, we won't know it.

Is your allowable "margin of error" also something that

could be describe as 0.5/5.5 or 9%, as the "relative"

margin of error?

Here's a stab at delineating the answer without being

sure of the question.

The distribution of responses, when it is a small fraction, can

be treated as a Poisson count.

For Poisson, the square-root is 'normalizing' such that

the result has a SD of 0.5.

- Or, SQRT(X) +/- 1.0 is the 95% CI .

- Or, you can name the sequential square-numbers to get

the 95% CI in the original metric.

Thus:

If you expect 100 actual responses, as a small fraction, the

95% CI is 81 to 121. (That is a relative margin of error

of +/- 20%.)

If you expect 400 responses, the CI is (381,421).

- If those definitions in the above match what you want,

it is simple algebra to extend work out the formulas for the

total Ns.

I imagine that you if you derive a formula from the above,

you should regard it as a crude estimate, until you see how

it behaves in practice.

The prediction makes it easier to spot divergences, but don't

be surprised if it performs badly, until you have monitored a few.

- How many changes are there in the mail-content

and in the audience between one mailing and the next?

--

http://www.pitt.edu/~wpilib/index.html

"Taxes are the price we pay for civilization." Justice Holmes.