A Comparison of Type I Error and Power Rates in Procedures Used Determining Test Dimensionality

Autor/inn/en	Guler, Gul; Cikrikci, Rahime Nukhet
Titel	A Comparison of Type I Error and Power Rates in Procedures Used Determining Test Dimensionality
Quelle	In: International Journal of Assessment Tools in Education, 9 (2022) 3, S.697-712 (16 Seiten) PDF als Volltext kostenfreie Datei Verfügbarkeit
Zusatzinformation	ORCID (Guler, Gul) ORCID (Cikrikci, Rahime Nukhet)
Sprache	englisch
Dokumenttyp	gedruckt; online; Zeitschriftenaufsatz
Schlagwörter	Comparative Analysis; Error of Measurement; Decision Making; Factor Analysis; Accuracy; Psychological Testing; Computer Software; Simulation; Sample Size; Evaluation Methods; Construct Validity + Suchen Sie Ihr Suchwort? Messfehler; Decision-making; Entscheidungsfindung; Faktorenanalyse; Psychological test; psychological tests; Psychological examination; Psychologischer Test; Simulation program; Simulationsprogramm
Abstract	The purpose of this study was to investigate the Type I Error findings and power rates of the methods used to determine dimensionality in unidimensional and bidimensional psychological constructs for various conditions (characteristic of the distribution, sample size, length of the test, and interdimensional correlation) and to examine the joint effect of the conditions (effect of the interaction of conditions) as well as the main effect of each condition. The simulative data were generated for the study using the SAS program. Within the scope of the study, the data were analyzed using the DIMTEST T statistic and the Dimensionality DETECT IDN index, which is one of the non-parametric methods. The Nonlinear Factor Analysis (NOHARM) method was preferred from among parametric methods. As a result of the study, it was noted that the most consistent results in making the unidimensionality decisions belong to the Nonlinear Factor Analysis method showing standard normal distribution according to the shape of the distribution. When the power study results were examined, it was noted that the DIMTEST T statistic gave more accurate results in conditions with large samples, consisting of data with standard normal distribution. On the other hand, while results of the DETECT IDN index and Nonlinear factor analysis were more internally consistent, it was noted that in conditions where the sample size was 1000 and above, the DIMTEST T statistic also made the right decisions in determining dimensionality. (As Provided).
Anmerkungen	International Journal of Assessment Tools in Education. Pamukkale University, Faculty of Education, Kinikli Campus, Denizli 20070, Turkey. e-mail: ijate.editor@gmail.com; Web site: https://ijate.net/index.php/ijate
Erfasst von	ERIC (Education Resources Information Center), Washington, DC
Update	2024/1/01