Monday, 3/26/01: Dr. Martin A. Koschat, Vice President, Research and Analysis, Time Inc. - Retail Sales and Marketing, AOL Time Warner, "News Vendors Tackle the News Vendor Problem"
Friday, 5/4/01: Professor Ka Sing Man, Syracuse University School of Management, "Long Memory Time Series and Short Term Forecasts"
Friday, 5/11/01: Professor Padmal Vitharana, Syracuse University School of Management, "Requirements Identification in Component-based Software Development: An Assessment Model"
Summary: The presentation will cover practical and strategic consideration in the distribution of consumer magazines. It will address the question of optimizing distribution under the established compensation arrangements in the channel, and it will address the question of how these compensation arrangements should be structured.
Regarding the practical aspect of the problem, in 1998 the Retail Sales and Marketing division of Time Inc. conducted a comprehensive review of the newsstand distribution principles and procedures for its portfolio of leading consumer magazines. This review resulted in changes in three major areas of distribution, including the evaluation of each magazineís national print order, the wholesaler allotment procedure, and the store distribution process. This revised process, referred to as Time Inc.ís Draw Management Program, has resulted in budgeted annual savings in excess of $3.5 mill.
Regarding the strategic aspects of compensation within the channel, it will be shown that the current system where each channel member receives a fixed fraction of gross revenue is, in general, sub-optimal, and an alternative revenue share arrangement between the channel members that Pareto-dominates the current arrangement will be presented.
PowerPoint file of presentation.
Friday, 5/4/01, 3:30-5:00, room 003 in the School of Management: Professor Ka Sing Man, Syracuse University School of Management, "Long Memory Time Series and Short Term Forecasts", refreshments provided
Summary: This talk begins with an introduction to time series with long memory behavior. Various examples including river-flow data, high frequency return and exchange rate data will be used for illustration. The idea of fractional differencing is popular in modeling the long memory characteristic. Linear model like ARIMA is extended to ARFIMA where the degree of differencing d is non-integer value, see for example Granger & Joyeux (1980), Hosking (1981). For modeling variance changes, FIGARCH, FISV and etc have been introduced, see for example Baillie et. al. (1996) and Bollerslev & Wright (2000).
In this talk, we will focus on studying the usefulness of low order ARMA model in the prediction of long memory time series with ARFIMA structure, where -0.5 < d < 0.5. In practice, Crato & Ray (1996) pointed out that the low success rate in the selection of the right ARFIMA model, along with the large variance in the estimation of the parameters, make the use of simple ARMA model appealing for forecasting purpose. We argue that if interest is in short term prediction, a suitably adapted ARMA(2,2) model can produce competitive forecasts. Theoretical results concerning the choice of the adapted parameters will be presented. Numerically, we find that its efficiency loss in predicting one step ahead is at most 0.6%, and at most 3% up to predicting 10 step ahead. However, caution needs to be taken when using the adapted ARMA model for long term prediction of strongly persistent time series. The predictability memory content of the adapted ARMA(2,2) model is also studied. For illustration, we forecast the US consumer price index for food. If time allowed, some related issues of estimating an ARMA(1,1) model with nearly cancelled parameters, and the relationship between ARFIMA process and Fractional Gaussian Noise will be discussed.
Friday, 5/11/01, 3:30-5:00, room 103 in the School of Management: Professor Padmal Vitharana, Syracuse University School of Management, "Requirements Identification in Component-based Software Development: An Assessment Model", refreshments provided
Summary: Requirements Analysis (RA) is the process of identifying usersí requirements or needs in determining what to build in a computer system. Software development literature is replete with studies that demonstrate how ineffective RA has led to failed applications. Some of the difficulties encountered in RA however are due to inherent limitations in traditional approach to software development. On the other hand, component-based software development (CBSD) presents a unique approach to developing software. Components advertise the services they offer and could be organized in a knowledge-base (i.e., repository). CBSD paradigm provides an effective communication vehicle for users and analysts by enabling them to uncover requirements as they navigate through the component knowledge-base.
In this paper, we draw from the information processing theory (IPT)
on problem solving to develop an assessment model for evaluating the impact
of CBSD on requirements identification, arguing that the access to components
in a knowledge-base facilitates the requirements identification.
The key elements of the IPT on problem solving are information processing
system of the problem solver, task environment, and internal representation
of the problem space. We propose that access to a component knowledge-base
enhances information processing system of the problem solver and simplifies
the task environment. This in turn clarifies userís internal representation
of the problem space and would result in generation of quality user requirements.
This theoretical framework makes it possible to empirically test the impact
of CBSD on requirements identification process.