measurement.txt

Set a criterion for a particular treatment or measure. This is a standard of measure that you can compare other values to/create a normative value or range Means of evaluating patients and their progress Compare and discriminate between individuals and groups Draw predictions about the predictive relationships between variables

Continuous and discrete. Continuous can adopt any number within a defined range (limited by the instrument measuring the variable), discrete variables must adopt a whole number that is specific to a category (heart rate)

• Nominal a scale in which there is no order and no classification, just a label. No quantitative value. (blood type, handedness, type of mental illness, side of hemiplegic involvement, etc)
• Ordinal Scale There is a defined order between the assigned individuals or objects, but there is no way of telling how much value is between them. (sensation scale = normal > impaired > absent)There may be a theoretical “zero” point, such as no palpable muscle contraction, or “negatives,” such as terrible-bad-ok-good-great, with ok being the zero point.
• Interval scale the interval scale is much like ordinal but with known and equal distances between the variables No true zero point, just an arbitrary chosen point that can descend into negative values (think celcius and farenheit, or calendar years)
• Ratio scale There is a true zero (meaning total absence of the quality being measured) and the ratings are empirical rather than arbitrary. Ratio data can be converted across scales (english to metric) and can be infinitely divided

• reliability - how good a testing procedure is at reproducing the same results every time (how dependable it is)
• validity - measuring what you intend to measure (knowing that your information matches the normative values of the literature)

• systematic error - a predictable and constant error (an incorrectly calibrated measuring instrument)
• randoms errors - fatigue, inattention, mechanical inaccuracy, or simple mistakes
• sampling error - sampling does not truly represent the population
• Regression toward the mean - extreme scores that are likely to contain error will regress toward the mean on subsequent trials
• reliability coeffecient - a coefficient closer to 1 is more reliable, a coefficient closer to zero is less reliable (T/(T+E))
• intrarater - how good a rater is at being dependable for consistent measurements
• interrater - how intuitive a test is for being able to be reproduced by multiple clinicians