Information Technology Reference
In-Depth Information
of computer platforms, demographic variables such as gender and ethnic-
ity, and the names of participants or the arbitrary code numbers assigned
to them.
Numerical coding of categorical variables is regularly done in the
course of data analysis. For example, in specifying the gender of the partici-
pants in a study, we would use numerical codes such as 1 and 2. Which
gender is associated with which numeric code is arbitrary and is not
important; what is important is that the coding scheme be used consis-
tently throughout the data set. Thus, some researchers always code
females as 1 and males as 2, whereas others may adopt a different coding
strategy for each separate data set.
2.1.3 ORDINAL SCALES
An ordinal scale of measurement uses numbers to convey “less than” and
“more than” information. This most commonly translates as rank order-
ing. Objects may be ranked in the order that they align themselves on some
quantitative dimension but it is not possible from the ranking informa-
tion to determine how far apart they are on the underlying dimension. For
example, if we were ranking the height of three people, a person seven feet
tall would be ranked 1, someone five feet, one inch tall would be ranked 2,
and someone else five feet tall would be ranked 3. From the ranked data,
we could not determine that two of the individuals were quite close in
height.
2.1.4 SUMMATIVE RESPONSE SCALES
Summative response scales, such as five-point and seven-point scales,
require respondents to assign numbers to represent attitudes or judg-
ments. For example, a five-point scale may have an anchor (a definition
of a scale point) for the value of 1 as
not very true for me
and an anchor for
the value of 5 as
very true for me.
Respondents would be presented with
a set of items (e.g., “I feel as though I am a worthwhile person,” “I enjoy
meeting new people at social gatherings”) and their task would be to place
the scale value that represented how true the statement was for them by
each item. This type of measurement can be traced back to Rensis Likert
(pronounced “lick-ert” by the man himself) in the 1930s (Likert, 1932;
Likert, Roslow, & Murphy, 1934; Murphy & Likert, 1937). Likert was try-
ing to find a more efficient way to accomplish what Louis Thurstone had
done in his pioneering work to develop interval level measurement scales
to assess attitudes (Thurstone, 1927, 1928, 1929; Thurstone & Chave,
1929).
The numbers on the scale are based on an underlying continuum
defined by the anchors and are ordered, typically in an ascending way,
to reflect more of the property being rated. It is called a
summative scale
because it is possible to add (summate) the ratings of several items on an
inventory and to divide by a constant (usually in the process of taking a
mean) to obtain an individual's score on the inventory.
