bei Google ScholarTitelangaben
Küsshauer, Alexander ; Schüler, Jens ; Baum, Matthias:
Test-Retest Reliability in Metric Conjoint Experiments : Important Requirement or Overrated Nuisance?
2020
Veranstaltung: 24. Interdisziplinäre Jahreskonferenz zu Entrepreneurship, Innovation und Mittelstand (G-Forum)
, 28.09. - 02.10.2020
, online.
(Veranstaltungsbeitrag: Kongress/Konferenz/Symposium/Tagung
,
Paper
)
Abstract
Within-subject experiments such as metric conjoint approaches are popular to study individual decision-making in the entrepreneurship literature. Taking a methodological perspective, the testretest reliability is usually treated as a key aspect that makes or breaks a study because low reliabilities may embody severe validity concerns. Traditional research suggested a threshold for internal consistency measures of r = 0.70, which have been adopted for test-retest reliabilities in within-subject designs. However, such a threshold might be inappropriate because conjoint
studies rely on several data points and thus maximize the reliability of results (Monsen, Patzelt, & Saxton, 2010). To address this critical puzzle, we first conduct a systematic literature review to evaluate how the current “rules of thumbs” may affect test-retest reliabilities. Second, we use primary data from a conjoint study on former entrepreneurs’ employability and simulate several data sets varying in sample sizes and test-retest reliabilities to analyze how regression estimates and model fits vary. Our results demonstrate that more nuanced perspectives toward test-retest reliabilities in within-subject designs are needed and best practice recommendations are offered.