The Structure of Objectivist Studies - Evaluation Methods in Biomedical Informatics - page 91

The set of terms identified by the system for each case is reviewed to see if
the case is indexed correctly. From a measurement perspective, name the
attribute and object class.
4. In studies of gene expression analysis, a DNA array (or chip) is used
to detect levels of messenger ribonucleic acid (mRNA) expressed in a bio-
logical sample (e.g., normal or cancer tissue cells). The array consists of
many different DNA strands of known sequence printed at defined posi-
tions; each such position has multiple copies of DNA strands that consti-
tute a coherent spot on the array. The biological mRNA sample is “reverse
transcribed” into complementary DNA (cDNA), labeled with a fluorescent
dye and hybridized (joined) with the DNA on the array. Only the cDNA
from the biological sample that matches the DNA strands on the chip will
be hybridized. Finally, the array is scanned and the amount of fluorescence
is assumed to correspond to the amount of hybridized DNA on each
spot. Viewing the chip as a measurement instrument, name the attribute(s)
being measured and the object class. Does this measurement process, as
described, include multiple independent observations?
[Answers to these questions are found at the end of the chapter.]
Levels of Measurement
Measurement assigns the value of an attribute to an object, but not all
attributes are created equal. Attributes differ according to how their values
are naturally expressed or represented. Attributes such as height and weight
are naturally expressed using continuous numerical values whereas attri-
butes such as “marital status” are expressed using discrete values. An
attribute's level of measurement denotes how its values can be represented.
It will be seen in later chapters that an attribute's level of measurement
directs the design of measurement instruments and the statistical analyses
that can be applied to the results of measurements performed on samples
of objects. There are four such levels of measurement.
1. Nominal: Measurement on a nominal attributes results in the assign-
ment of each object to a specific category. The categories themselves do not
form a continuum or have a meaningful order. Examples of attributes mea-
sured at the nominal level are ethnicity, medical specialty, and the bases
comprising a nucleotide. To represent the results of a nominal measurement
quantitatively, the results must be assigned arbitrary codes (e.g., 1 for
“internists,” 2 for “surgeons,” 3 for “family practitioners”). The only aspect
of importance for such codes is that they be employed consistently to
represent measurement results. Their actual numerical or alphanumerical
values have no significance.
2. Ordinal: Measurement at the ordinal level also results in assignment
of objects to categories, but the categories have some meaningful order or