Re: "fixing" a control

Were cases randomly assigned to test and control?

How were the segments chosen? Are they exhaustive of the strata of the pop of interest? A random set of values, e.g., places?

If the outcome is a continuous variable, might have a two way ANOVA. # of segments by test/control?

If the outcome is a dichotomy, you might have a contingency table?

Art
Art@xxxxxxxxxxxxx
Social Research Consultants
Inside the Washington, DC beltway.



b_branford@xxxxxxxxx wrote: 
Hi,

Thanks for your response. The context is direct marketing campaigns. We
send out promotional offers to a "test" population and then compare
their behaviors to the "control" population to see if we have a lift.

Actually, the problem is in the way the mail plan was designed it is
difficult to have an overall "test" and overall "control".

Step 1: Segment 1 was selected based on certain criteria and then
broken up into "Test" and "Control".

Step 2: Segment 2 was created (after suppressing segment 1) and then
broken up into "test" and "control".

.....etc.

Obviously, if you combine all the test cells and all the control cells
together, they will not look similar (comparison of means, etc).
However, for analytical purposes it would be useful to have an overall
"test" population and an overall "control" population which look
similar. Yes, I see your point regarding how do we compare. In fact,
that is another question I had -- is there a multivariate way of
validating that a test population looks similar to a control or, if
not, is there a way to do it at the backend?

Thoughts ?

Thanks.

brad



Data Matter wrote:

The bad news for you is you can't "fix" the control cell.  I presume
the experiment has been completed so there is no way to re-run history.

The most important thing for you to find out is why did the control
cell look different.  Was it a problem with the randomization
procedure?  Is it a systematic problem?  Were your sample sizes small?

What do you mean by "not looking alike"?  Are you comparing means?
variances?  are you comparing the response variable or other variable?

Any post-hoc "fix" would involve a multitude of assumptions, which will
likely invalidate your conclusion anyway.  Without further details, it
is impossible for us to provide useful advice.



b_branford@xxxxxxxxx wrote:

hi,

I have a control cell which does not look like the test cell. Any
thoughts on how I can "fix" the control so that it looks like the test
population. 

thx.

brad 


.

Relevant Pages

  • Re: "fixing" a control
    ... You can either compare the test and control groups based on the ... variables used to segment the population and/or the response ... >>> I have a control cell which does not look like the test cell. ...
    (sci.stat.edu)
  • THE EPIGENETIC SYSTEM OF HEREDITY
    ... CNS control of the deposition of maternal cytoplasmic factors in the egg cell ... This explains the fact that some parts of the embryo (those ... genes) but by the epigenetic information maternally deposited in the egg in the ...
    (sci.bio.evolution)
  • tell Zakariya its exact specifying beside a formation
    ... information structures the balance. ... A single cell, in its DNA strands, holds a MASSIVE AMOUNT OF INFORMATION. ... ground in a muddy puddle of our constituent chemicals. ... It can yield the ULTIMATE in control. ...
    (sci.crypt)
  • Re: Simple? Query the contents of an Imagebox
    ... the images within the control ... cannot compare the image to it's original value. ... Public Function CompareTwoImages(ByRef image1 As Image, ... Dim img1 As New Bitmap ...
    (microsoft.public.dotnet.languages.vb)
  • Re: Text box date format
    ... I'm not sure of the interaction you want from the TextBox back to the control source cell, but assuming you only want to get the American date from the cell when you display the UserForm for use in the code associated with the UserForm, try this... ...
    (microsoft.public.excel.programming)