Re: Mixed 2-way ANOVA

On 28 Sep 2005 03:50:14 -0700, "Corky" <corky.abc@xxxxxxxxxxxx> wrote:

> Hi,
>
> I am looking to carry out an experiment, i want to compare a control
> group made up of all particpants (to obtain a baseline) to 3 other
> groups (made up of all the particpants split into 3 groups -
> participants will be put into these groups depending on set criteria -
> the groups will probably be unequal). Each of the 4 groups will carry
> out 2 different tests and three measures will be collected.

I agree with Bruce's observation, that this seems to be 3 by
<something> instead of 4x2.

I want to further mention that when participants are "put into
these groups depending on set criteria", you apparently throw
away the chance of drawing the relatively-firm conclusions that
are possible for Randomized trials. Or did you intend some
other meaning?

In my consulting experience, a lot of people settled for two-group
designs, because it takes a lot of cases to get good power.
Even though there would be reasons why it would be neat to
compare 3 or 4 or 8 groups, it does take more cases.

If you have one-control versus many-trials, then your best
power is achieved by assigning extra cases to the Control.

But, of course, if you are not Randomizing, then you are
left with a contentious argument at the end, anyway, about
whatever you are testing.... One solution might be, call
it all "exploratory."


>
> So I will have a 4X2 mixed design. Can anybody help me out with sample
> sizes, i.e. how many participants will i need for each of the 4
> conditions in order for it to be statistically significant. I have read
> pages on power anylsis, but i t seems i need to carry out a pilot
> study, for which i don't have time. Any ideas?

One of the best guidelines, all the time, is to ask,
"How many cases are usually used for similar questions?"
- then, figure they probably did not use enough, and
add 50% or so.

If it is not randomized, you will want to see rather large
within-subject effects, even if the Ns can be large; since
you will have to contend that "other factors" could not have
been responsible for the outcome.


--
Rich Ulrich, wpilib@xxxxxxxx
http://www.pitt.edu/~wpilib/index.html
.

Relevant Pages

  • Re: secret codes embedded in your printer output...
    ... >In every society you have those who exercise power and they generally ... introduce any law they wish with few constraints. ... and where there is no solid constitution and BoR to supervise them. ... >But I think there is an agenda to exert more and more control over ...
    (uk.politics.misc)
  • Iranian Power Grab Debacle (a primer)
    ... positions of power over who and what lives or, more likely, dies. ... be unleashed upon Iran. ... control over, ... The USA gathers petroleum from around ...
    (alt.politics)
  • Iranian Power Grab Debacle (a primer)
    ... positions of power over who and what lives or, more likely, dies. ... be unleashed upon Iran. ... control over, ... The USA gathers petroleum from around ...
    (alt.politics.bush)
  • Re: Declaration of Separation
    ... We do not seek to control anything. ... Regardless of the parties in power, ... longer accept unearned suffering as our role in life. ... realm -although we demand that they do no damage to our realm. ...
    (soc.men)
  • Declaration of Separation
    ... We do not seek to control anything. ... Regardless of the parties in power, ... longer accept unearned suffering as our role in life. ... realm -although we demand that they do no damage to our realm. ...
    (soc.men)