Multicollinearity in Marketing Models: Notes on the Application of Ridge Trace Estimation in Structural Equation Modelling pp3‑15
Abstract: Multicollinearity in Structural Equation Modelling (SEM) is often overlooked by marketing scholars. This is unfortunate as multicollinearity may lead to fallacious path coefficient estimates or even bring about statistical non‑significance of the parameter estimates. Previous empirical illustrations on mitigating the effects of multicollinearity are virtually non‑existent in the literature. The purpose of this paper is to empirically illustrate the problem of multicollinearity in marketing mod els and the use of ridge trace estimation in mitigating the effects of multicollinearity in SEM, using the LISREL program. Two slightly differing ridge estimation procedures are illustrated using real data with a multicollinearity problem: Method A, in wh ich the ridge constant is added manually to all diagonal elements of the correlation matrix of the variables in the model, and Method B, in which the ridge constant is added manually only to the diagonal elements of the correlation matrix of the exogenous and explanatory endogenous variables in the model. In evaluating suitable values of the ridge constant, the ridge trace method is used. It is concluded that ridge trace estimation is an effective way of mitigating the effects of multicollinearity in SEM. With same ridge constant values, both methods produce same point estimates of path coefficients, but Method B produces smaller standard errors of parameter estimates and larger squared multiple correlations than Method A.
Keywords: marketing modelling, multicollinearity, structural equation modelling, ridge trace estimation, LISREL
Abstract: The normality assumption behind ANOVA and other parametric methods implies not only mound shape, symmetry, and zero excess kurtosis, but also that data are equidistant. This paper uses a simulation approach to explore the impact of non‑equidista nce on the performance of statistical methods commonly used to compare locations across several groups. These include the one‑way ANOVA and its robust alternatives, the Brown‑Forsythe test, and the Welch test. We show that non‑equidistance does affect the se methods with respect to both significance level and power, but the impact differs between the methods. In general, the ANOVA is less sensitive to non‑equidistance than the other two methods are and should therefore be the primary choice when analyzing potentially non‑equidistant data.
Abstract: IS research has matured significantly over the last three decades, leading to increasingly complex research designs as well as complex analytical techniques to analyze the data collected. Similar advances have happened in the experimental and qu asi‑experimental designs. Some key characteristics of these advances are: 1) use of latent variables approaches to operationalize key variables; 2) the need to understand the causal relationship between elements of the study; 3) the need to study the e ffects of technology as an addition to existing methods of working; and, 4) recognition that some conditions create greater change in outcomes over time. In spite of these advances in data collection and design, researchers are still confirming data coll ected via experiments to use ANOVA for analysis. This paper outlines an analytical technique that moves Information Systems experimental research beyond ANOVA. By combining and extending three advances in Structural Equation Modeling techniques, namely Me an and Covariance Structure analysis, Stacked Group modeling and Latent Growth modeling, the paper outlines a robust analysis technique that accommodates the above‑mentioned advances in experimental design. The technique provides for an in‑depth test of a ll model assumptions, as well as the flexibility to accommodate an increasing variety of experimental designs. A detailed example is provided to illustrate the usage of the technique in an Information Systems context. The example shows now only the accomm odations needed in an information systems context, but also how this technique can be used to extract results from existing research methods that was not possible with ANOVA. The arguments presented in the paper as well as the example on how to use should provide future researchers with a guideline on how to use these techniques as well as provide a platform for how they can extend these techniques to accommodate more research method advances.
Keywords: Keywords: SEM, experiments, stacked group modeling, latent growth modeling, invariance, Information Systems
Abstract: There is wide agreement in the methodology community that the choice of data collection mode may affect the quality of response. In addition, the method of choice may also influence respondent behavior and feelings which may also impact the qual ity of data. This research examines response quality and respondent satisfaction measures compared across three data collection methods among male adolescent respondents. Results suggest that adolescents provide improved levels for several dimensions of response quality when participating in interview‑based research studies compared to text‑based methods such as electronic form and paper and pencil formats.
Keywords: Keywords: Response quality, adolescent respondents, male respondents, interview-based methodology, text-based methodology, response distortion
Using the Multiple Case Study Design to Decipher Contextual Leadership Behaviors in Indian Organizations pp54‑65
Abstract: This paper demonstrates how the complex phenomenon of contextual leadership in business organizations was studied in a unique manner by using the multiple case study design. In the current context of fast paced change, uncertainty and ambiguity, leadership roles in organizations assume great significance. Recent studies have indicated the relevance and importance of studying leadership behavior in the context in which they appear and not away from it. In this study, the multiple case study desig n was used for the twin purposes of capturing rich descriptive contexts of the leader and strengthening the patterns of findings using Yins (1984) replication logic.Within the case studies, mixed methods were employed to generate qualitative and quant itative data simultaneously on the contextual leadership behaviors of senior Indian managerial leaders. The methodology,based on the social phenomenology paradigm, used interviews to capture the interpretation of the leaders about their environments. Qual itative data was collected through interviews, company documents, industry reports and analysts reports. Quantitative data collection methods included a scale based on Ansoffs model, the adaptive capacity scale as well as the Multifactor Leadership Quest ionnaire. The study proposes a model of leadership based on rich synthesis of patterns of leadership behavior across contexts in an emerging markets scenario using the multiple case study design, mixed methods in data collection and analysis, a combinat ion of data driven and theory driven codes in the coding framework and mixed methods for transforming the raw dataThe objective of this study was to provide insights into designing a multiple case study research and carrying out cross case analysis using matrices. Additionally the study describes the usage of the multiple case study design to study leadership embedded in its context in a novel manner.
Abstract: The appropriateness of case studies as a tool for theory testing is still a controversial issue, and discussions about the weaknesses of such research designs have previously taken precedence over those about its strengths. The purpose of the pa per is to examine and revive the approach of theory testing using case studies, including the associated research goal, analysis, and generalisability. We argue that research designs for theory testing using case studies differ from theory‑building case s tudy research designs because different research projects serve different purposes and follow different research paths.