Gender differences in technology adoption - application of PLS-MGA

• Published in: Statisztikai szemle 2017-01, Vol.95 (1), p.51-77
• Main Authors: Nagy, Judit T, Bernschutz, Maria
• Format: Article
• Language: Hungarian
• Subjects: Bootstrap method; Gender differences; Statistical analysis; Technology adoption
• ISSN: 0039-0690
• DOI: 10.20311/stat2017.01.hu0051

A latens valtozos modellezes soran nem ritka, hogy heterogen megfigyelesekkel van dolgunk, melynek figyelmen kivul hagyasa problemat okozhat az elemzes folyaman. A tanulmany olyan technikat mutat be, ami a PLS-SEM (partial least squares structural equation modeling - parcialis legkisebb negyzetek strukturalis egyenletek modellje) alkalmazasakor kinal megoldast a megfigyelt heterogenitas figyelembevetelere. Ezen az un. MGA-n (multi group analysis - tobbcsoportos elemzesi eljaras) tobb modszer is alapul, melyek kozul a szerzok a parameteres probakat, a permutaciotesztet, a Henseler-fele MGA-modszert, valamint a nemparameteres konfidencia-intervallumok modszeret tekintik at. Ezt kovetoen az utobbi harom eljaras alkalmazasat egy oktatasi adatsoron mutatjak be. A vizsgalati modell alapjat a technologia elfogadasanak modellje szolgaltatja.//The PLS-MGA methods are suitable for comparing the path-coefficients of two predetermined groups in the course of PLS path analysis. The parametric tests are the modified versions of the independent sample t-test referring to the path-coefficients, which - in contrast to the PLS-SEM basic method - require normality, and therefore their usage needs further investigation. The non-parametric NGA methods, however, do not require any distributional assumptions. The permutation test is a randomization method, which only requires the nearly identical size of group-specific samples. For the comparison of the group-specific path coefficients, Henseler's MGA method uses the bootstrap distribution, while the non-parametric bootstrap confidence interval method uses the bootstrap confidence intervals of path coefficients. There are several methods applied for setting up the bootstrap confidence intervals. From these, the authors demonstrated the simple bootstrap percentile method (Bp method), which in the case of a small sample cannot be applied properly, as well as the double-bootstrap percentile method (double-Bp method) based on double bootstrap sampling that is suitable for making more precise confidence intervals. [web URL: http://www.ksh.hu/statszemle_archivum#year=2017/issue=01.] Reprinted by permission of Statisztikai Szemle