دانلود رایگان ترجمه مقاله روش های اندازه گیری و اعتبارسنجی در سیستم اطلاعات مدیریت (نشریه Jstor 2011)

اندازه گیری ساختار و روش های ارزیابی در MIS و پژوهشی رفتاری: تلفیق فنون فعلی و جدید

Construct Measurement and Validation Procedures in MIS and Behavioral Research: Integrating New and Existing Techniques

Despite the fact that validating the measures of constructs is critical to building cumulative knowledge in MIS and the behavioral sciences, the process of scale development and validation continues to be a challenging activity. Undoubtedly, part of the problem is that many of the scale development procedures advocated in the literature are limited by the fact that they (1) fail to adequately discuss how to develop appropriate conceptual definitions of the focal construct, (2) often fail to properly specify the measurement model that relates the latent construct to its indicators, and (3) underutilize techniques that provide evidence that the set of items used to represent the focal construct actually measures what it purports to measure. Therefore, the purpose of the present paper is to integrate new and existing techniques into a comprehensive set of recommendations that can be used to give researchers in MIS and the behavioral sciences a framework for developing valid measures. First, we briefly elaborate upon some of the limitations of current scale development practices. Following this, we discuss each of the steps in the scale development process while paying particular attention to the differences that are required when one is attempting to develop scales for constructs with formative indicators as opposed to constructs with reflective indicators. Finally, we discuss several things that should be done after the initial development of a scale to examine its generalizability and to enhance its usefulness.

1 Introduction

It has been over 20 years since Straub (1989) made the following observation about the MIS literature:

Instrument validation has been inadequately addressed in MIS research. Only a few researchers have devoted serious attention to measurement issues over the last few decades…and while the desirability of verifying findings through internal validity checks has been argued by Jarvenpaa, et al. (1984), the primary and prior value of instrument validation has yet to be widely recognized (p. 147).

Approximately a dozen years later, in a retrospective on the Straub article, Boudreau et al. (2001) surveyed the MIS literature again to assess whether there had been any improvement in the use of construct validation techniques, and concluded that their “findings suggest that the field has advanced in many areas, but, overall, it appears that a majority of published studies are still not sufficiently validating their instruments.” (p. 1). Similar concerns regarding the practices used to validate constructs have also been expressed in the field of management by Scandura and Williams (2000), who compared the methodological practices reported in three top journals in two different time periods (1985–1987 and 1995– 1997), and concluded that there had actually been a decrease in the proportion of studies that reported information about construct validity and reports of discriminant, convergent, and predictive validity. Therefore, the observation that Bagozzi and Phillips (1982, p. 468) made over 25 years ago still rings true: “Scientists have found it very difficult to translate this seemingly simple notion [of construct validity] into operational terms.”

The reason for the apparent lack of progress in this area certainly is not due to a shortage of articles written on the technical procedures that should be used to validate scales (e.g., Anderson and Gerbing 1988; Anderson et al. 1987; Bagozzi et al. 1991; Diamantopoulos and Winklhofer 2001; Edwards 2001; Fornell and Larcker 1981; Gerbing and Anderson 1988; Nunnally and Bernstein 1994; Straub et al. 2004). However, one possibility is that the researchers reading these articles absorb only a portion of what is said because many of these articles are complex and require a fairly well-developed technical knowledge of structural equation modeling procedures. The result is that readers may not understand how to implement the recommendations made in these articles. An even more likely possibility is that there is simply so much work on the topic of scale development and evaluation that it is difficult for researchers to prioritize what needs to be done. Indeed, we believe that one reason Churchill’s (1979) seminal article has proven to be so useful to researchers is that he outlined an organized set of activities that set priorities for what needs to be done in the scale development and evaluation process. Therefore, in the spirit of Churchill, the goal of this research is to provide an updated set of recommendations that can be used to give researchers a framework for developing valid scales.

We believe that there are several reasons why an updated set of recommendations would be useful. First, many of the scale development procedures advocated in the literature fail to adequately discuss how to develop appropriate conceptual definitions of a focal construct. Second, many of the recommendations are based on an improper specification of the measurement model2 that relates the latent variable representing a construct to its measures.3 Finally, techniques that provide evidence that the scale actually measures what it purports to measure have been underutilized in the management and MIS literatures. In the sections that follow, we will briefly elaborate on each of the limitations identified above. Following this, we will discuss each of the steps in the scale development process while paying particular attention to the differences that are required when one is attempting to develop scales for constructs with formative indicators as opposed to constructs with reflective indicators. Finally, we discuss several steps that should be taken after the initial development of a scale to examine its generalizability and to enhance its usefulness.