Many educators would not agree with the initial assumption that number of
examination passes is indicative of a school’s academic excellence. However, if you
accept this contentious idea then agreement with statement C and the measurement
assumption of individual ranked scores would seem most appropriate. Many educators,
because of the problems of construct validity, would be happier with statement D because
it makes the fewest measurement assumptions. It does, however, also tell us the least
about the schools.
2.3 Practical Decisions about MeasurementShould an underlying continuum exist theoretically, it does not necessarily mean that it
can be measured. Such measurement depends upon the availability of a suitably sensitive
or refined measurement device. For example, common constructs such as attitudes and
motives may be measured by scoring individual responses on a rating scale, from say 1 to
- Here it would be assumed that individuals vary continuously along a continuum but on
which one can not make direct refined measurements. Discrete measurements are
therefore used by which individuals score either 1 or 2 or 3, etc. It is further assumed that
individuals who score, 2 on a 5-point attitude scale are distinguishable only by the fact
that subjects having a similar quantity of the attribute also score 2 and these subjects
differ in amount of attitude from those who score 1, 3, 4 or 5.
Measurement assumptions for psychological and educational tests and scales are
different from those appropriate to physical measures such as height and weight
(measured on ratio scales). These different measurement assumptions lead to different
conclusions. For example, zero marks on a test of science knowledge may not mean that
the person tested has no science knowledge (probably the test is inappropriate for that
individual). It would, however, be valid to conclude that a measurement of zero metres
means no height. Whereas it would be valid to conclude that the difference between a
child that is 1.0m tall and another that is 1.25m tall is the same as the difference between
a child that is 0.75m tall and another that is 1.0m tall, it may not be valid to conclude that
a person who scores 4 on a 5 point attitude scale has twice as much attitude compared to
a person who scores only 2 on the same scale.
Different measurement assumptions, for example, the distinction between discrete and
continuous measurements, are often over emphasized and certainly are not always clear.
This ambiguity between discrete and continuous measures seldom matters given that the
same statistical methods can often be used for both types of measurement, particularly if
the discrete measurement scale has fine gradations and care is taken when interpreting the
results.
Attaining at least interval levels of measurement for tests and scales is a measurement
and interpretation problem and not a statistical one. Statistics deals with numbers whereas
the researcher has to deal with the properties underlying any numerical measurements.
For example, interval level measures are often tacitly assumed, even when it is obvious
that it is not realistic to do so. Subsequent statistical analysis may then yield numbers
which in themselves are numerically correct. It is the consumers of statistical data, who
Measurement issues 25