شیوه ها و عملکرد توسعه محصول :تجزیه و تحلیل معادله ساختاری مبتنی بر مدل سازی چند گروهی
کد مقاله | سال انتشار | تعداد صفحات مقاله انگلیسی |
---|---|---|
2695 | 2006 | 22 صفحه PDF |
Publisher : Elsevier - Science Direct (الزویر - ساینس دایرکت)
Journal : International Journal of Production Economics, Volume 103, Issue 1, September 2006, Pages 286–307 Cover image
چکیده انگلیسی
This paper provides an illustration of the strategies associated with conducting tests of model invariance that are uncommonly applied and reported in the Operations Management literature. The illustration uses a theoretical model that describes the relationships between key product development practices (i.e., heavy-weight product development managers, information technology use and concurrent engineering) and performance as measured through product innovation and quality. Based on a sample of 214 manufacturing executives and managers, the study considers a priori proposed measurement and structural models and further examines the extent to which these models are invariant across two different cellular manufacturing environments.
مقدمه انگلیسی
The overwhelming majority of recent empirical research in the field of Operations Management employs some type of structural equation modeling (SEM) analysis. More often than not, however, there is little consideration that the given sample examined in the research study may be comprised of multiple groups. This is particularly important as the components of a proposed model (e.g., the so-called measurement and/or structural components of a model) may not always be invariant across the different groups. For example, certain manufacturing practices may only improve performance in large firms but not in small firms. While running separate models for large and small firms may provide some insight into potential differences between employed manufacturing practices, a more sophisticated statistical testing approach within SEM must be undertaken in order to pinpoint specific similarities and discrepancies. In the SEM methodological literature, general statistical tests dealing with hypotheses about potential group differences are commonly referred to as tests of model invariance (Heck and Marcoulides, 1989; Marcoulides and Heck, 1993). The terms interaction modeling or multi-sampling are also often used to refer to data analyses that aim to compare the similarities of proposed models across different samples or subgroups of samples (Schumacker and Marcoulides, 1998). Of course, all these modeling techniques are really just part of the more general class of approaches encompassed by SEM (Raykov and Marcoulides, 2000). The analysis basically begins by fitting a model to the data for each sample considered separately with none of the parameters constrained to be equal across groups. This unconstrained model serves as the baseline model. Subsequently, in a stepwise fashion, more stringent constraints are placed on the model by specifying the parameters of interest to be constrained across groups. The model is then examined using a chi-square (χ2) difference test between the less restrictive and more restrictive models to determine whether the model and the individual parameter estimates (e.g., factor loadings, factor inter-correlations, error variance, structural relations) are invariant across the samples. A significant difference in χ2 represents a deterioration of the model and the null hypothesis that the parameters are equal is rejected. A non-significant χ2 difference is consistent with model invariance; that is, the parameters examined are equal across groups. The above tests of model invariance also represent an approach to construct validation, which essentially demands that if the field of Operations Management is to advance as an academic discipline, greater attention must be paid to ensure that models developed for a given environment are also examined in other environments. Consequently, the crucial and decisive question is ultimately whether constructs developed to measure certain operations management principles and practices are invariant across different environments of interest. According to Horn and McArdle (1992), measurement invariance basically refers to “whether or not, under different conditions of observing and studying phenomena, measurement operations yield measures of the same attribute (p. 117)”. Obviously the consequences of not establishing invariance can be detrimental to the integrity of any research findings. For example, differences in scale means for levels of JIT practices might be due to true differences between small and large firms or due to systematic biases in the way people from large and small firms respond to certain items. Differences in relationships between constructs as exhibited among large and small firms could indicate real disparities in structural relations between constructs as scaling artifacts, disparities in scale reliability, or even non-equivalence of the constructs involved. The purpose of this paper is to provide a non-technical overview of strategies associated with conducting tests of model invariance that are uncommonly applied and reported in the Operations Management literature. The paper is written more from a didactic perspective so as to serve as a guide for researchers interested in applying this methodology but who may be uncertain of the strategies involved. To accomplish this purpose, we focus on examining the invariance of a model across only two types of firms: (i) those that have a high cellular manufacturing presence, and (ii) those that have a low cellular manufacturing presence. Case and industry studies focusing primarily on large, highly visible firms have substantially improved our understanding of key product development practices (e.g., heavy-weight product development managers and concurrent engineering) and describe how these practices impact product innovation and quality (Donnellon, 1993; Millson et al., 1992; Koufteros et al., 2001, Koufteros et al., 2002b and Koufteros et al., 2005). An essential question is whether these practices actually contribute equally in different environments. While firms have adopted new and innovative product development practices, many have also adopted innovative manufacturing practices such as cellular manufacturing. Cellular practices are purported to improve operational efficiencies by producing one family of parts within each cell. Because parts produced in each cell are “similar,” setup times are reduced. Because similar products are made within each cell, quality may improve as well. Beyond operational efficiencies, however, it is conceivable that the relationships among product development practices and performance would be different under varying levels of cellular manufacturing. Perhaps firms with heavier presence of cellular manufacturing would be more accommodating to product development efforts and thus boost the ability of firms to introduce new products and features. This research study describes a multivariate model for examining relationships among key product development practices (i.e., heavy-weight product development managers, information technology use, concurrent engineering), and performance such as product innovation and quality. Beyond the prototypical strictly confirmatory testing of an a priori proposed model, multi-group analyses are carried out both on the measurement and structural parts of the model. The two groups are formed based on specified levels of cellular manufacturing. The hypotheses are tested with SEM via the LISREL program on a sample of 214 manufacturing firms. The study makes an important contribution to the field of Operations Management in important ways. First, given the timely need for applied researchers to comprehend and adopt the mechanisms involved in testing model invariance across multiple groups, the paper will clearly help in this instructional process. There appears to be a lack of such applications in the literature to date, and thus the paper could provide readers with a useful reference tool to follow when conducting their own research. The study also contributes to the body of Operations Management literature with substantive findings. For example, the effects of concurrent engineering on quality were found to be non-equivalent across cellular manufacturing levels.
نتیجه گیری انگلیسی
The inclusion and study of measurement or structural model invariance is essentially absent in the Operations Management literature. Of course, in general, this is not unique to just the Operations Management field. SteenKamp and Baumgartner (1998) provide the following six reasons for the lack of research using invariance methods in the marketing literature: “(1) the bewildering array of types of measurement invariance that can be found in the literature, (2) the lack of an agreed upon terminology to refer to the different kinds of measurement equivalence, (3) researcher's relative unfamiliarity with testing measurement models that incorporate the latent and observed variable means, (4) the considerable methodological complexities involved in testing for different kinds of measurement invariance, (5) researcher's uncertainty about the extent to which measures have to be equivalent in order for particular comparisons to be meaningful, and (6) the absence of clear guidelines as to how to ascertain whether or not a measure exhibits adequate invariance (p. 79)”. Despite these concerns, testing for (measurement or structural) model invariance is an important methodology for addressing some difficult research questions in various contexts. Beyond illustrating how one can test a measurement and structural model, this paper presents a general paradigm for conducting multi-group analyses and provides an overview of some of the most salient tests that are carried out in a multi-group data analysis. Of course, it is important to note that the paradigm presented is one of many that exist in the literature (e.g., Vandenberg and Lance, 2000; Bensaou et al., 1999; Steenkamp and Baumgartner, 1998). And, although there are some subtle differences between the various paradigms that one may encounter in the literature, these differences are typically found in the sequencing or order in which the tests take place. The flowchart presented in Fig. 3 illustrates a step-by-step paradigm through which the measurement model invariance is first tested. This first test is then followed by structural model invariance tests and/or latent mean invariance tests, depending on the theoretical question a researcher is positing: in this case, an investigation of a theoretical model that relates product development practices and competitive capabilities. Sheremata (2000) proposes that “successful development requires structures and processes that generate and retrieve new ideas, knowledge, and information and then integrate this intellectual material into collective action” (p. 390). Organizations can be described as dynamic systems that require balance and equilibrium in order to maintain momentum and proceed toward their goals. Sheremata describes two forces that when in balance, organization action is collective as well as creative leading organizations to find high-quality solutions to problems quickly and efficiently. Centripetal forces are structural elements and processes that integrate dispersed ideas, knowledge, and information into collective action. On the other hand, centrifugal forces make the ideas, knowledge, and information available that organizations need for creative action. The model presented in this paper includes both centripetal as well as centrifugal forces. The model provides for the free flow of information by team members, the individuals closest to the source of the problem. The free flow of information and data is enabled by the use of computers. There is also a high relational density in the network of product development actors as they belong to a cross-functional team that begins its work from the early stages of product development and works concurrently through all the phases. Heavy-weight product development managers are influence agents and catalysts who act pro-actively and who can motivate others to individual and collective action. The results presented in this study indicate that firms use several integrated product development practices (i.e., heavy-weight product development managers, concurrent engineering, and computer utilization). These practices help firms cope with uncertainty and equivocality in the environment by providing a champion in the product development process, improving knowledge sharing through teams/concurrent engineering, and increasing information availability and flow. These practices seem to be central to the enhancement of product innovation and quality. Analysis of the structural model indicated that heavy-weight product development managers do have overall significant, direct, and positive effects on the use of concurrent engineering. Heavy-weight product development managers have the expertise and positional authority to champion product development efforts including marshaling resources that are required to support concurrent engineering efforts. Heavy-weight managers communicate effectively with top management in a language they understand, and instill confidence that bottom-line results can be improved as computer-based product development efforts are supported. Their role as agents of integration is vital. Although indirect effects were not specifically examined, Table 3 suggests that heavy-weight product development managers do have significant indirect effects on product innovation and quality. The far-reaching effects of heavy-weight product development managers cannot be ignored. The cornerstone for concurrent engineering is integration. This stimulates information gathering and processing. These tools increase the concurrency of design by allowing teams of designers to communicate across an integrated computer network, thereby lessening the need to co-locate participants. This environment links project participants through an information system that enables the exchange of knowledge, establishes easy and effective mechanisms to coordinate activities, and shortens product development time. The environment has transparent links among the various computer and software systems, a common storage and access mechanism, and user interfaces which allow all disciplines to access the design tools. Information technology tools provide the capabilities to capture past practices, which can be indexed and cross-referenced for fast and easy access by other product development teams. They increase the firm's ability to manage multiple product development projects and to provide feedback and direction, quickly. These outcomes can also be used for interactive training that brings new team members up-to-speed, rapidly. As these tools become standardized through use across the firm, there is substantial flexibility in making assignments to product development teams because participants become to some extent “interchangeable parts.” These knowledge-sharing activities enable firms to shorten product development time, increase productivity, and improve training opportunities. The customer, the organization, and its key suppliers can be electronically connected so design decisions are made quickly and implementation begins immediately. The use of computer information technology tools can also increase standardization of product development practices. This enhances flexibility because engineers and managers can be moved from one project to another without excessive start-up costs. This cross-fertilization enhances learning and can lead to the rapid spread of innovative ideas and approaches. The use of computers is associated with higher reported levels of concurrent engineering. The effects of IT use on concurrent engineering were invariant across cellular manufacturing environments. The use of computers may also be indirectly related to product innovation and quality as the results presented in Table 3 indicate. Concurrent engineering did exhibit significant direct effects on product innovation and quality. The indirect effects on quality may also be significant as product innovation apparently mediates the relationship. The effects of concurrent engineering on quality appear to be different across cellular manufacturing levels as Table 4 illustrates. In fact, the effects are found to be stronger in a low cellular manufacturing environment. This came as somewhat of a surprise as we expected it would be the other way around. The effects of concurrent engineering on product innovation, however, were not found to be different. The effects were found to be equally potent in both environments. High levels of product innovation are associated with high levels of quality, which supports the prevalent view in the strategy literature. Although the results support this notion, the path coefficients are actually different across cellular manufacturing levels. In fact, the effects are quite stronger in the high cellular manufacturing environment. There are some obvious limitations to our study. Foremost amongst them is that in this study we only examined internal integration in product development. Nevertheless, it is essential that external integration, which has become an important aspect of product development and supply chain management, also be investigated in future studies. As product development teams achieve internal integration, they recognize the need for and seek external integration by forming strategic partnerships involving customers and suppliers to coordinate activities across the value chain. Internal integration is an antecedent, and perhaps a determinant, of a firm's effort to achieve external integration. Computer usage enables cross-functional information sharing and links the firm's problem solving efforts with customers and suppliers. Early involvement of functional specialists and team-based problem solving efforts create an information rich environment. In this context, it becomes apparent that many internal problems require a coordinated cross-functional solution that demands information sharing and joint problem solving with external constituents. We clearly recognize that such variables are important and, to the extent that they can be identified and operationalized, they may contribute to either the prediction or moderation of proposed theories. Nonetheless, this paper illustrates an important tool that can be used for evaluating the generalizability of proposed models in the field of Operations Management across potentially different groups and environments.