Re: nomogram
- From: glenbarnett@xxxxxxxxxxxxx
- Date: 16 Jul 2006 20:22:30 -0700
Natalie Vivien wrote:
If so, how do we use the nomogram to look for the corresponding sample
size? The chart is the one which has two scales by the sides (one for
difference and one for power) and a straight line in the middle with a
decreasing gradient (indicating the sample size)?
I assume the nomogram you have is an appropriate one for your
experimental design.
Most nomographs are 3-scale alignment charts.
It sounds like you specifically have an N-chart:
| \ |
| \ |
| \ |
| \ |
| \ |
| \ |
| \ |
| \ |
| \ |
| \ |
To use one, you need a straightedge**.
**A suitable straightedge like a ruler, or a clear strip of plastic
with a /thin/ line marked on it, or even a piece of thread stretched
taut.
Find the two values you know (in your case, the difference and the
beta) on their scales. Lay your straightedge so it joins those two
values. Where it crosses the remaining scale, that's your required
sample size. If the value is obscured by your straightedge, touch a
fine tipped pencil (say a mechanical pencil) to the scale where it
crosses and then remove the straightedge, so you can see where the tip
of the pencil is. (If it's between values, you'll want to round up the
sample size to get at least the desired power.)
I hope this helps.
One neat thing about nomograms (besides being a lot faster and less
error-prone than calculator-punching) is that you don't need to do any
algebra to solve for a different variable. If you know difference and
sample size, you can read your beta straight off its scale.
Glen
.
- References:
- nomogram
- From: Natalie Vivien
- nomogram
- Prev by Date: Re: piecewise constant approximation
- Next by Date: Differing Scope and Methodology in GAO and University of California Reports Account for Variations in Cost Estimates for Homosexual Conduct Policy. GAO
- Previous by thread: Re: nomogram
- Next by thread: Isoplot
- Index(es):
Relevant Pages
|
|
