定量研究方法的评价-四个旅游相关的案例
本文旨在通过对酒店业的旅游涉入程度、工作投入与工作满意度之间的关系进行研究。数据来自台湾20家国际酒店的336名员工，并通过结构方程模型分析。这篇文章强调了共同方法偏差，它用一个共同methodfactor方法统计控制方法偏差。探索性因素分析的结果表明，所有的项目在目前的研究中加载了四个不同的因素，占大多数变量之间的协方差没有主导因素加载。因此，有人说，常见的方法偏差不是本文中的重大问题。
人们普遍认为，常见的方法偏压膨胀由自我报告测量的变量之间的关系。Becauseself-报告显然适合于工作满意度的研究，在本文中作者使用自我报告收集数据。因此，常见的方法偏差都在这里进行分析。总体而言，笔者避免一些因素会导致共同方法偏差。首先，它从不同的来源得到的预测和标准变量的措施。例如，工作投入的数据来自员工和工作满意度的数据和来自酒店管理人员。其次，本文对参与者的匿名性和保密性进行了研究。再次，该问题的顺序的平衡是确保在这篇文章。第四，所用的量表项目有明确和准确，编写，因此较少受到偏见影响。
This article aims to exam the relationship among tourism involvement, work engagement and job satisfaction in the hotel industry. The data come from 336 employees of 20 international hotels in Taiwan and was analysed via structural equation modeling. This article stressed the common method bias and it used a singlecommon methodfactor approach to statistically control for method biases. The outcomes of an exploratory factor analysis demonstrated that all of the items in the current study were loaded on four different factors and no dominant factor accounted for the majority of the covariance among the variables. Hence, it said that common method bias is not the major problem in this article.
It is widely argued that common method bias inflated relationship between variables measured by self-reports (Podsakoff et al., 2003). Becauseself-reports are clearly appropriate for job satisfaction research, in this article the author use self-report to collect data (Chan, 2009).Therefore, common method bias are analysed here. Overall, the author avoided some factors that can cause common method bias. Firstly, it obtained measures of predictor and criterion variables from different sources. For example, the data of work engagement were from the employees and the data of job satisfaction were from the hotel managers. Secondly, this article mentioned that it ensured the anonymity and confidentiality of the participants. Thirdly, the counterbalance of the order of the questions is ensured in this article. Fourthly, the scale items it used are written clearly and precisely and, so less subject to bias.
In terms of test method, this article use singlecommon methodfactor approach to test whether it has common method bias. Commonly, a single methods factor is an appropriate representation for the phenomenon, especially for examining common-rater effects such as social desirability (Siemsen et al., 2010). However, Meada (2007) argued that the use of a single factor test can only provide an indication of whether a single factor accounts for all the covariance among the items, this process actually does nothing to statistically control for method effects.In this article, four factors emerged from the factor analysis, this can not proof that the measures are free of common method variance. Indeed, if it can, then it would mean that common method variance would have to completely account for the covariance among the items because it to be regarded as a problem in this study. Therefore, single common method factor approach may be not a useful remedy to deal with the problem so attention can be paid to other remedies are better suited for the purpose such as particle correlation technique. Moreover, using combine methods and continuum of ways to statistically control are suggested.
In addition, conceptual overlap in items used to measure different constructs can cause relationships bias (Conway et al., 2010). However, this article does not provide all items for its measures, so it is hard to figure out if it has conceptual overlap in items.
Article 2 Robin N. et al. (2012), Residents’ support for tourism: An Identity Perspective, Annals of Tourism Research, Issue 1, January 2012, Pages 243-268This study aims to analyse community support in the island economy of Mauritius by testing a model based on the social exchange theory and the identity theory. This study collected data from the residents of Mauritius. Specifically, a random sampling was used to select the pre-determined number of respondents from different occupational identity, environmental identity, and gender identity of the residents. In this study, attitude played the role of mediator and Sobel z test was used to test the mediating effects. The result support mediating effects of attitude to the negative influences in the relationship between gender identity and support for tourism.
Holmbeck (1997) argued the hypothesis that a predictor is related to the mediator should have a clear theoretical rationale. The hypotheses in this article are based on social exchange theory so thetheoretical support for the mediators is enough. In this study, attitude has similar relation with gender identity and support for tourism, which is also desirable for a mediator analysis in choosing mediators because the mediators’ relations with the predictors and outcomes are similar(Frazier et al., 2004).
In this article, the author used Sobel z test to analyse the mediator effects because it is considered to havegreater statistical power than other formal methods of assessing mediation. However, Baron (1986) indicated that the mediated effects were large this method only when sample sizes were large enough (larger than 500). In this study, the sample size is only 300 so it can not support that the mediated effects were large in this study by choosing this analysis method.
Besides, in general, SEM is considered the preferred method (MacKinnon,2002). One of the advantages of SEM is that it can control for measurement error, provides information on the degree of fit of the entire model, and is much more flexible than regression. (Frazier et al., 2004) Also, the authors did not test whether Sobel z test it employed was better fit the data than other alternative models.
Article 3 Ben H. et al. (2013), culture conflicts and culture cushion?Annals of Tourism Research, Volume 43, Pages 321-349This article explores antecedents of perceived discrimination of tourists, which is an under-researched area in tourism. A questionnaire survey was conducted with248 Chinese and 240 Caucasian tourists in Hong Kong. This study highlighted the moderating role of intellectual competence and power distance. It used multiple regression analysis to test the moderators. The results are showed in table 1.The results indicate that the effect of power distance is significant in equation 3 and the change in F-statistics between equations on the main and conditional effects as well as the interaction terms are significant in equation 4.Hence, the moderating effects of intercultural competence and power distance are supported.
Table 1 Moderated Regression Results for Intercultural Competence and Power Distance on Anticipated Discrimination
Moderator variable Equation 1 Equation2 Equation 3 Equation 4
Intercultural
competence _0.08(0.08) _0.12(0.08)
Power distance 0.20(0.07)*** 0.35(0.10)***
*** P < 0.01.; ** p < 0.05.; * p < 0.10.
This article looked for moderators to fill the theoretical gap of anticipated discrimination because there are unexpectedly weak relations between culture and discrimination in the previous study. First of all, the moderator analysis this article employed is grounded in the theory of Attribution theory. When designing a study to test moderation, theory is very important (Frazier et al., 2004).This article used an appropriate designed empirical research. Also, the sample size of 488 is also enough to achieve adequate power. Because the moderator variables in this article are not continuous variables, the regression analysis method that this article is applied is also generally (Frazier et al., 2004). |