Page 153 -

120 Part 2 • InformatIon requIrements analysIs

MeasuReMent. Systems analysts commonly use two different forms of measurement scales:
1. Nominal scales
2. Interval scales
Nominal scales are used to classify things. A question such as the following uses a nominal scale:

What type of software do you use the most?
1 5 A word processor
2 5 A spreadsheet
3 5 A database
4 5 An email program
Obviously, nominal scales are the weakest forms of measurement. Generally, all an analyst can
do with them is obtain totals for each classification.
In interval scales, the intervals between each pair of numbers are equal. Due to this charac-
teristic, mathematical operations can be performed on the questionnaire data, resulting in a more
complete analysis. Examples of interval scales are the Fahrenheit and Celsius scales, which measure
temperature.
The foregoing example of the information center definitely does not use an interval scale,
but by anchoring the scale on either end, the analyst may want to assume that the respondent
perceives the intervals to be equal:

How useful is the support given by the Technical Support Group?
Not Useful at All Extremely Useful
1 2 3 4 5

If the systems analyst makes this assumption, more quantitative analysis is possible.
validity and ReliaBility. There are two measures of performance in constructing scales:
validity and reliability. A systems analyst should be aware of these concerns.
Validity is the degree to which a question measures what the analyst intends for it to mea-
sure. For example, if the purpose of the questionnaire is to determine whether the organization is
ready for a major change in computer operations, do the questions measure that?
Reliability measures consistency. If a questionnaire was administered once and then again under
the same conditions and if the same results were obtained both times, the instrument is said to have
external consistency. If the questionnaire contains subparts and these parts have equivalent results, the
instrument is said to have internal consistency. Both external and internal consistency are important.

constRucting scales. The actual construction of scales is a serious task. Careless construction
of scales can result in the following problems:
1. Leniency
2. Central tendency
3. Halo effect

Leniency is a problem caused by respondents who are easy raters. A systems analyst can
avoid the problem of leniency by moving the “average” category to the left (or right) of center.
Central tendency is a problem that occurs when respondents rate everything as average.
The analyst can improve the scale (1) by making the differences smaller at the two ends, (2) by
adjusting the strength of the descriptors, or (3) by creating a scale with more points.
The halo effect is a problem that arises when the impression formed in one question carries
into the next question. For example, if you are rating an employee about whom you have a very
favorable impression, you may give a high rating in every category or trait, regardless of whether
it is a strong point of the employee’s. The solution is to place one trait and several employees on
each page rather than one employee and several traits on a page.

Designing Questionnaires
Many of the same principles that are relevant to the design of forms for data input (as cov-
ered in Chapter 12) are important in designing questionnaires as well. Although the intent of a