A psychometric evaluation of the measurement invariance and measurement equivalence of the basic traits inventory (BTI)

Date
2019-12
Journal Title
Journal ISSN
Volume Title
Publisher
Stellenbosch : Stellenbosch University
Abstract
ENGLISH SUMMARY : The main objective of an organisation is to generate profit through the products and services it delivers to the market. Organisations can only exist if there are people that enable the organisation to function and to deliver the services and/or products to market. It is therefore in the organisations' best interest to look after their human resources and to ensure that this is a priority throughout the whole organization. The primary objective of the human resource function is to enhance the performance of the employees in the organisation in a manner that adds optimal value to all functions of the organisation. The human resource function involves human resource interventions to fulfil this objective. Through recruitment and selection, the human resource function aims to select and place applicants, with the best possible probability of achieving success, within the organisation. Recruitment and selection should be free from all forms of direct and indirect unfair discrimination as stated in the Employment Equity Act (EEA) 55 of 1998, to ensure that all candidates that can do the work equally well have the same chance of succeeding in the application process. The EEA also prohibits the use of biased psychological tests and other similar assessments. The objective of this research study was to determine whether gender bias exists in the Basic Traits Inventory (BTI) using measurement invariance and measurement equivalence as an approach, as outlined by Dunbar, Theron and Spangenberg (2011). A sample of archival item response data for South African men and South African women that completed the English version of the BTI was made available to the researcher by the developers of the BTI. The single-group BTI measurement model in which the latent first-order dimensions were operationalised via the individual items of the instrument was first fitted to samples of male and female data separately. The null hypothesis of close fit was not rejected for the single-group male and single group female measurement models. Using multi-group confirmatory factor analysis, a series of hierarchical multi-group measurement models on which increasingly stringent equality constraints were imposed were then fitted, starting with configural invariance as the first step. Configural and weak invariance were confirmed. However, metric equivalence could not be confirmed. Partial metric equivalence was consequently tested for and confirmed. Partial weak strong invariance was confirmed. Scalar equivalence could again not be confirmed. The data analysis results revealed that partial scalar equivalence could be confirmed for the BTI multi-group measurement model. Partial weak partial strong strict invariance was confirmed. Conditional probability equivalence could, however, again not be confirmed. Partial conditional equivalence could be confirmed. Non-invariant items were highlighted in the study. More specifically six non-uniform biased items, 108 uniform biased and 120 error variance biased items were identified. This research study concludes with a discussion of the results found, limitations of the study and recommendations for future research.
AFRIKAANSE OPSOMMING : Organisasies bestaan om 'n wins te maak deur middel van die dienste en/of produkte wat dit aan die mark lewer. Organisasies kan slegs bestaan indien daar mense is wat die organisasie in staat stel om te kan funksioneer en om die dienste en/of produkte te lewer. Dit is dus in organisasies se beste belang om na hul menslike hulpbronne om te sien en om te verseker dat dit 'n prioriteit is in die hele organisasie. Die hoofdoelwit van die menslike hulpbronfunksie is om die prestasie van die werknemers in die organisasie te verhoog op ’n wyse wat optimaal waarde toevoeg tot al die dele van die organisasie. Hierdie doelwit word nagejaag deur menslike hulpbronintervensies. Deur werwing en seleksie poog die menslike hulpbronbestuurfunksie om mense te keur en te plaas wat die beste moontlike kans op sukses sal hê in die organisasie. Die Wet op Diensbillikheid (WDB) 55 van 1998 stipuleer dat werwing en keuring vry van enige vorm van direkte of indirekte onbillike diskriminasie moet wees om alle kandidate wat die werk ewe goed kan doen ‘n gelyke kans te gee om die pos te bekom. Die WDB verbied ook die gebruik van sydige sielkundige toetse en ander soortgelyke instrumente. Die doel van hierdie navorsingsstudie was om geslagsydigheid in die Basic Traits Inventory (BTI) te bepaal, deur metings-invariansie en metingsekwivalensie as ’n benadering te gebruik soos voorgestel deur Dunbar, Theron and Spangenberg (2011). 'n Steekproef argivale item responsdata van mans en vroue wat die BTI voltooi het, is aan die navorser beskikbaar gestel deur die ontwikkelaars van die BTI. Die enkelgroep BTI metingsmodel waarin die latente eerste-orde persoonlikheidsdimensies geoperasionaliseer is deur die individuele items van die instrument is ten aanvang afsonderlik op die manlike en vroulike steekproewe data gepas. Die nul hipotese van benaderende passing is nie verwerp vir die manlike en vroulike enkelgroep metingsmodelle nie. Deur middel van die gebruik van multi-groep bevestigende faktorontleding is ’n reeks hierargiese multi-groep metingsmodelle gepas waarop toenemend veeleisende gelykheidsbeperkinge afgedwing is, beginnende met konfigurale invariansie as eerste stap. Die navorsingstudie het konfigurale en swak invariansie bevestig. Metriese ekwivalensie kon egter nie bevestig word nie. Daar is toe getoets vir gedeeltelike metriese ekwivalensie, en dit is bevestig. Gedeeltelike swak sterk invariansie is bevestig. Skalare ekwivalensie kon egter nie bevestig word nie. Daar is egter steun gevind vir gedeeltelike skalare ekwivalense. Gedeeltelike swak gedeeltelike sterk streng invariansie is bevestig. Voorwaardelike waarskynlikheid ekwivalensie kon egter weereens nie bevestig word nie. Gedeeltelike voorwaardelike waarskynlikheid ekwivalensie kon wel bevestig word. Die nie-invariante items is uitgewys tydens die studie. Meer spesifiek is ses nie-univormig sydige items, 108 univormig sydige items en 120 foutvariansie sydige items geïdentifiseer. Laastens, om die navorsingstudie af te sluit, word die navorsingsresultate bespreek, beperkinge word uitgewys en voorstelle vir verdere navorsing gemaak.
Description
Thesis (MCom)--Stellenbosch University, 2019.
Keywords
Basic traits inventory (BTI) -- South Africa, Personality questionnaires -- Statistical methods -- South Africa, Errors-in-variables models, Measurement bias, Measurement invariance, UCTD
Citation