Databases Reference
In-Depth Information
Kirs (1993) and Kim and March (1995); details controlled in these experiments include
trainer characteristics and instructional examples. Table 1 summarizes the variables used
in prior research.
Key Findings from Prior Studies
The results from the empirical studies reviewed can be categorized as follows: a)
Effects of
data modeling formalism
on
user performance and attitudes,
b) Effects of
user
characteristics on user performance and attitudes,
and c) Effects of
task characteristics
on
user performance and attitudes
. Most of the studies have focused on the fi rst category. In
addition to the associations between research variables, we will review the results for various
task components (facets) and the main lessons from the studies with a process focus.
Effects of data modeling formalism on user performance and attitudes.
The studies that
have investigated the effects of the data modeling formalism on performance and attitudes
can be divided into the following subcategories: a) those comparing a semantic model to
the relational model, b) those comparing two semantic models to each other, and c) those
comparing a semantic model with object-oriented models.
In the fi rst subcategory, the seven studies (Amer, 1993; Batra & Antony, 1994; Batra et
al., 1990; Jarvenpaa & Machesky, 1989; Juhn & Naumann, 1985; Liao & Palvia, 2000; Sinha
& Vessey, 1999) that have investigated the differences between the ER/EER and relational
modeling formalisms have all found support for the positive effect of the use of the ER/EER
model on one or several aspects of modeling performance. The studies provide strongest
support to ER/EER's advantage in modeling binary 1:M and binary M:N relationships; four
of the studies (Amer, 1993; Batra et al., 1990; Liao & Palvia, 2000; Sinha & Vessey, 1999)
support this fi nding, whereas the other fi ndings related to the identifi cation of relationships
and cardinalities, faster learning, understanding the notation, modeling ternary 1:M:N and
unary relationships, and generalization modeling are all based on only one of the studies.
For the binary relationships, these results are in line with those of Cao, Nah, and Siau's
(2000) meta-analysis, which included both modeling and query writing studies; our analy-
sis did not fi nd the same strong support for ER/EER's advantage over relational model in
modeling ternary 1:M:N relationships as theirs did. The one study (Shoval & Even-Chaime,
1987) that focused on the relationship between the relational model and a non-ER semantic
model, NIAM, found the relational model to lead to better user performance and to require
less time. As to the effects of the modeling formalism choice between semantic and rela-
tional models and the user attitudes, the results are scarce and inconclusive: Jarvenpaa and
Machesky (1989) found that subjects perceived the ER/EER model to be easier to use than
the relational model, but Shoval and Even-Chaime (1987) found that the subjects preferred
the relational model over NIAM.
Six studies (Hardgrave & Dalal, 1995; Lee & Choi, 1998; Liao & Palvia, 2000; Shoval
& Frumermann, 1994; Shoval & Shiran, 1997; Sinha & Vessey, 1999) have investigated
the effects of the choice between object-oriented models (although not consistently the
same ones) and ER/EER. The lack of consistency between the studies makes it diffi cult to
draw any general conclusions, but the direction of the studies seems to suggest that using
the ER/EER model leads to better performance in modeling tasks. The studies together
indicate that the use of ER/EER has a positive effect on modeling performance in fi ve of
the modeling facets (unary 1:1, binary 1:1 and 1:M, and ternary 1:M:N, and M:N:O), but,
unfortunately, the fi ndings come from different studies that do not provide support for
