Management Science in Practice by Unknown

Author:Unknown
Language: eng
Format: epub

Figure 7.2 Bias and precision

This all, of course, assumes we’re actually measuring the right thing – which brings us to our third and fourth attributes. In practice, we’re often not looking at an aspect that has an independent existence that we can touch or actually measure. Instead, many of the aspects we would like to use in our analyses we cannot measure directly; we have to use an approximate measure, or indeed a proxy measure (where we measure something that we feel is representative of the aspect in which we’re interested).

The third attribute of a measure is its reliability, which essentially refers to how consistent it would be if it measured the same aspect more than once. Obviously, precision affects this (a wildly imprecise measure would not give the same results if used more than once), and the effect of lack of precision needs to be removed from our consideration of reliability – imprecision refers to the attempt to measure something directly when our measuring instrument is not able to capture precisely the correct number, while lack of reliability refers to the attempt to measure something with the best instrument that we can but we are not sure it always measures the right aspect. The use of a ruler, for example, would be pretty well fully reliable for measuring a short length, but not necessarily precise.

Lack of reliability can come from a variety of sources, and we need to consider each of these when considering the reliability of a measure. The intrinsic reliability of measures can be investigated by simple repeated measuring, assuming a measurement does not change the measure (for example, if you repeat an IQ test immediately, memory of the previous test might affect the second test) or the time between tests does not change the answers. Instruments such as questionnaires measuring a concept can have multiple questions pertaining to that concept, and the results to the different questions can be compared. Reliability relating to the measurer can be studied by more than one person administering the test to similar groups that could be expected to have similar scores (this can help to reduce the problem of measurers being more likely to see what they expect to see). And so on.

This does all assume we are actually measuring what we think we are measuring. The fourth attribute of a measure is validity. A measure is valid if it really does measure what we are trying to measure – in other words, if the measurements we are achieving do tell us something valid about the aspect we think we are measuring. Sophisticated studies are sometimes carried out which find interesting results about a measure, but where the measure doesn’t actually represent the aspect in which the investigators were interested.

Let’s think about practical examples:

•Take for example productivity. You might measure the resources used by a factory (money, manpower etc.) over the space of a month, and the output on that month, and you might feel that the ratio of one to the other gave a good indication of productivity.

Download

Copyright Disclaimer:
This site does not store any files on its server. We only index and link to content provided by other sites. Please contact the content providers to delete copyright contents if any and email us, we'll remove relevant links or contents immediately.