Date:         Tue, 27 Sep 2005 23:59:17 -0400
Reply-To:     Richard Ristow <wrristow@mindspring.com>
Sender:       "SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>
From:         Richard Ristow <wrristow@mindspring.com>
Subject:      Re: need help on 2 different samples
Comments: To: Roy FGH <roykrim94@gmail.com>
Comments: cc: Marta García-Granero <biostatistics@TERRA.ES>
In-Reply-To:  <1d11d57305092706472e7ca0dc@mail.gmail.com>
Content-Type: text/plain; charset="iso-8859-1"; format=flowed

I'm building from Marta's reply:

>At 09:47 AM 9/27/2005, Roy FGH wrote:
>
>>RF> the situation is : i want to explain that attempts on increasing
>>the
>>RF> motivation of worker, e.g. raise the salary (var A, measured in
>>60
>>RF> workers) will result in better service that customers get (var B,
>>RF> measured in 30 costumers). Martha has guessed it, and advanced my
>>RF> question. Thank you, Martha.
>>RF> but i still wondering, is there any other statistical operation i
>>can
>>RF> use to answer my hypothesis, except correlation and regression of
>>RF> course. or maybe you can show me the best data collecting and
>>RF> processing technique so i can test the hypothesis.
>
>At 10:12 AM 9/27/2005, Marta García-Granero wrote:
>
>Unless you can link a customers to workers, I see no solution with
>your present data collection and layout.
>
>If you had measures of the customers service before and after the
>salary raise, you could try to compare means.
>
>Any other thoughts from other listers? (c'mon, one step forward, don't
>let all the fun for me...).

Marta, one does not lightly step forward when you've addressed a
statistical methods question.

But, Roy: You wrote, "is there any other statistical operation i can
use?".

Try dropping thoughts of statistical methods, for the moment.
Statistical methods won't find in data, what isn't there. And in most
simple cases, anything the statistics can see, you can see. (In these
cases, the statistical methods are mainly to keep you from seeing
things that are not there.)

So, think of two cases: Suppose that raising the salaries doesn't
change customer service even a little bit. Or, suppose that raising
salaries makes service to customers hugely, enormously better.

Now: How would you expect your data to look differently in those two
cases? If you had no statistics, but just had to look; and if you knew
that one of those was true; how would you decide which?

Answering that kind of question is the prelude to even thinking about
statistical methods.