Premium
Computing normative ranges without recruiting normal subjects
Author(s) -
Yaar Israel
Publication year - 1997
Publication title -
muscle and nerve
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.025
H-Index - 145
eISSN - 1097-4598
pISSN - 0148-639X
DOI - 10.1002/(sici)1097-4598(199712)20:12<1510::aid-mus5>3.0.co;2-b
Subject(s) - normative , goodness of fit , statistics , normal values , range (aeronautics) , gaussian , normal distribution , mathematics , reference values , audiology , medicine , psychology , physics , philosophy , materials science , epistemology , quantum mechanics , composite material
Abstract Computing normative data by recruiting normal subjects is extremely difficult. However, many who are examined in a typical clinical neurophysiology lab are normal. In this study we show how to use this abundance of referred subjects to compute normative distal latency statistics from the values themselves. If all latencies are displayed on a frequency distribution, the very left side, the shorter latencies, belong to the left side of the Gaussian “bell” of normal subjects. By curve‐fitting that side one can compute the coefficients of the latter. We started with an initial range of 2.0–3.6 ms and then recursively added data points until a “goodness‐of‐fit” criterion maximized. We computed these coefficients from 982 median motor distal latencies, showing highly significant fit ( P < 0.00001): mean at 3.76 ms and SD of 0.45 ms. The data analyzed in this study are only an example of the technique. The results are unique in that they were derived mathematically, without using normal subjects. © 1997 John Wiley & Sons, Inc. Muscle Nerve 20: 1510–1514, 1997