Re: Power Analysis and Multiple Regression and Indifference

Ben-Bard...

In additiokn to the comments by Reef Fish (I agree with those comments)
I want to add a little more. Your covariates have explained about 81%
of the variation in "home transaction values". That leaves about 19%
of the variation to work with. So you are hoping that "envionmental
factor" will explain a signficiant amount of that remaining variation.
You didn't mention the t-ratios associated with the 20 or so
covariates. Are they notably large or are most of them marginal?

There's one glaring hazard here. That "environmental factor" may be
correlated with one or more of the covariates, so you may have already
"consumed" the information re" the environmental factor. Get the entire
correlation matrix and look for large correlations among the factors
yoiu are considering. OMU



Ben-Bard wrote:
I am hoping someone can describe a method that will allow me to
establish an appropriate sample size for the test I am
undertaking...Any advice would be very appreciated.

I will be constructing a test where I am trying to determine if home
transaction values are affected by an environmental condition it can
see. A fairly small percentage (15-20%) of the sample of homes will
have a view of the condition. I will first need to control for all
other characteristics of the home, such as number of acres, square
feet, type of exterior (roughly 20 controlled variables). In so doing
I have received an adjusted r-squared of roughly .81. Then to the
model I add the variables of interest, distance from and visibility of
the condition. I hope to find that view of the condition is
statistically significant, and has its confidence intervals inside the
zone of indifference (aka it is equivalent to zero). The zone of
indifference will be +/- 3 or 4%.

I have experimented with PASS, and have not found the correct construct
to do a power analysis. I suspect it is there, but I have not found
it.

Any Ideas?

other info: Sample mean $102,000, SD $52,000, Covaried Standard Dev
$25,000.

.