Dear Allstaters,
I wish to summarize the output of a clinical test which provides a status map
of a patient's field of vision. Each of the 50-odd examined test points is given
a symbol which refers to the p-value from comparison with a proprietary
reference distribution from healthy subjects (<0.05, <0.02, <0.01, <0.005).
Until now I have simply scored these p-values using a simple, arbitrary scale
(p>5%=0; p<0.05=1; p<2%=2; etc.) and then summed them up to obtain an overall
ordinal 'abnormality' score.
It occurs to me that a better way to do this might be to assign a score based on
some transform of the p-value (eg. a probit or logit). For example, p<5%=2.2;
p<2%=3, p<1%=3.9; etc -- which presumably would preserve more of the
information.
Does anyone have any thoughts on this? I would also be grateful for any hints on
potentially useful references. I appreciate your help.
Paul
Assistant Prof (Ophth Vis Sci)
Dalhousie University, Halifax, CANADA
|