Information Technology Reference
In-Depth Information
from these theories to validate the labels for the factors. This matching of factors
to concepts is a data-driven interpretation of the factors (see Sect.
16.1.2
) and might
bring us closer to a “correct” interpretation of a factor.
In Fig.
16.4
, we show the cumulative score the observers gave the concepts for
each factor. The colour coding in this figure indicates how much the observers
agreed that the concept could explain the factor: the dark-coloured cells indicate
unanimous agreement, the light-coloured cells indicate that three out of the four
observers agreed. The initial factor analysis interpretation of the factors is indicated
with an asterisk.
The fit of the factors and the concepts determines the validity of the interpreted
theory for this factor. The observers unanimously matched most factors to the
concepts corresponding with the initial factor analysis interpretation of the factor;
see Fig.
16.4
in which the asterisks indicate the initial interpretation and the dark
cells indicate unanimous matching. The factors where the observers disagreed
(not unanimous) with the initial factor interpretation are police factors 1, 9, 11, and
13 and suspect factors 1, 3, and 10 (see Fig.
16.4
).
We see several explanations for this disagreement; first, the initial subjective
interpretation of the factors might have been wrong. The factors with a higher
number had fewer items loading on them (and less explanatory power), which might
have made it more difficult to interpret them. Four of these higher factors (suspect
factors 9, 11, and 13 and police factor 8) had few and diverse items loading on
them, which resulted in disagreements during the initial factor interpretation. It is
likely the disagreement persisted in the current analysis for suspect factors 9, 11,
13 and police factor 10. Second, factors could initially have been interpreted as
having an “absence of something”. This was the case for police factor 1 which
was initially interpreted as “missing rapport”. In the subsequent “mapping concepts
to factors” task, the observers did not unanimously match the concepts of rapport
to police factor 1. This might be because the instructions were unclear what to
do when a concept was explicitly absent: some observers said that this factor
contains information about the concept rapport (i.e. rapport is missing) while others
said there is no rapport so the concept of rapport is not present. For suspect factor
3 and 10 we can give no alternative explanation and conclude that our initial factor
interpretation was incorrect (see Table
16.5
).
16.4.3
Relations Between Theories
The observers unanimously matched factors to concepts from theories that were
not initially included in the factor analysis. In other words, more concepts than
initially come to mind might play a role in explaining a factor. For example, suspect
factor 4 was interpreted during factor analysis as a
together
stance, but the factor
was not only matched to the concept
together
(from theory of stance), but also
to
positive approval
(from the theory of face), and
attention
and
positivity
(from
rapport) (see Fig.
16.4
).
Search WWH ::
Custom Search