Need Question Answered
eo at cbnewsb.cb.att.com
Fri May 14 10:23:28 EST 1993
In article <1993May10.210403.4926 at alw.nih.gov> arm at helix.nih.gov (Andrew Mitz) writes:
>In article <9305080126.AA59210 at itsa.ucsf.EDU> jharper at ITSA.UCSF.EDU (M. Jane Harper) writes:
>>On 6 May 93 ltorre at eis.calstate.edu asked:
>>>Can someone out there help me distinguish between the scientific terms:
>>"Accuracy" is not a technical measurement term which, to my knowledge, has a
>>meaning outside its common one. "Reliability" means the extent to which an
>>instrument reproduces results given the same analysand. In other words, if
>>you and I (or Machine A and Machine B) rate someone (or some thing) on a
>>characteristic and we get identical or closely correlated results, the tool
>>we used to perform the rating has a high reliability. "Validity" is a related
>>concept which refers to the extent to which an instrument measures what it is
>>intended to measure. If I stick a thermometer in someone's mouth, I will get
>>a temperature, not a blood pressure; the thermometer is a valid measure of
>>temperature and an invalid measure of blood pressure.
>Good answer. Let me add that sometimes Accuracy is used to distinguish from
>precision. In this case, precision means the same as "reliability" as defined
>above. Accuracy would mean the difference between a measurement and the
>actual value being measured. Rarely is the actual value known, but sometimes
>it is. For example, if I have 100 pennies (measured by counting), and I use
>a device that measures the number of pennies by using their weight, I could
>test the accuracy of the device.
Yes, accuracy and precision are often incorrectly considered symonyms.
Your definition of accuracy is correct, but precision is not "reliability,"
it is "refinement." For example, take two sticks of equal length. Graduate
stick A into 10 equal intervals and stick B into 20 equal intervals.
Measurements taken with stick B will be more "precise" than those with
stick A. However, is either stick an "accurate" measure of inches?
As for reliability and validity, the psychologist George Kelly had these
definitions of them when used to describe psychological tests:
Reliability is a measure of a test's insensitivity to change.
Validity is a measure of a test's ability to tell us something we already know.
(These definitions make more sense when you know how the reliability and the
validity of a psychological test are determined.)
Ed Oliveri, eo at cbnewsb.att.com
More information about the Neur-sci