Model specification issues in PLS-SEM: Illustrating linear and non-linear models in hospitality services context

Abstract

Purpose

Prior research in tourism and hospitality domain calls for closer attention to model specification when using partial least square-structural equation modeling (PLS-SEM), including the choice of software and algorithm for PLS model estimation. This paper aims to illustrate the significance of choosing appropriate algorithms for testing the nature of relationships by comparing findings using two different PLS-SEM software packages.

Design/methodology/approach

Using a field experiment, relationships between visual servicescape aesthetics and affective responses are conceptualized based on literature in environmental psychology and marketing domains. With photographic surrogates as stimuli in two different hospitality service contexts – spa and upscale restaurant – data are collected from 350 respondents.

Findings

By comparing results of SmartPLS 3.2 and WarpPLS 5.0 software and theoretical understanding from environmental psychology literature, it is illustrated that the results and their interpretations may not be in line with theory if model specifications are not correctly implemented and are not addressed through usage of software with a relevant algorithm to test them.

Originality/value

The study highlights the implications for model specification issues such as type of variables and nature of relationships that tourism and hospitality researchers often face and also how use of appropriate algorithms can overcome limitations of model testing for complex models and provide empirical rigor to support theory.

研究目的

本论文使用两种不同的PLS-SEM处理软件来测试理论模型。通过解析模型设定参数问题, 特别是通过结构关系本性分析, 本论文指出选择合适的软件测试模型在酒店旅游领域的PLS研究中是非常关键的。

研究设计/方法/途径

本论文借助图像拍摄手段采用实验的采样方式, 在两个不同的酒店服务场所—按摩和高档餐厅—搜集350份数据。本论文采用Smart PLS 3.2 和Warp PLS 5.0 软件来测试PLS-SEM。 这两款软件支持线性和非线性理论关系的比较。

研究结果

通过Smart PLS 3.2 和Warp PLS 5.0 软件得出的报告结果分析, 不同软件处理PLS得出的结果可能有偏差, 而且会不符合理论设定。如果模型设定参数不正确, 通过使用合适的PLS-SEM软件和相关的数据分析加以辅助, 可能会解决参数不正确的问题。

研究实践意义

本论文的比较分析结果可能会帮助到酒店和旅游领域的研究者们, 在做对有关PLS-SEM软件选择的时候, 哪些软件可以更加合适的测试模型有着参考意义。

研究原创性/价值

本论文重点指出了模型设定参数的相关问题, 比如旅游酒店领域常见的变量种类和关系属性等。本论文还研究了如何选择合适的数据分析方法来克服测试复杂模型时的限制, 并且提供实践结果来支撑理论。

Citation

Kumar, D.S. and Purani, K. (2018), "Model specification issues in PLS-SEM: Illustrating linear and non-linear models in hospitality services context", Journal of Hospitality and Tourism Technology, Vol. 9 No. 3, pp. 338-353. https://doi.org/10.1108/JHTT-09-2017-0105