Eduardo Saliby

Descriptive sampling: an improvement over Latin hypercube sampling

Descriptive Sampling (DS), a Monte Carlo sampling technique based on a deterministic selection of the input values and their random permutation, represents a deep conceptual change on how to carry out a Monte Carlo application. Abandoning the paradigm that a random selection of sample values would be necessary in order to describe random behavior, DS is a rather polemical idea. An interesting issue related to DS are the similarities between it and Latin Hypercube Sampling (LHS) to be discussed in this paper. After a brief description of both methods, it is shown how close DS and LHS are. As such, DS can be seen as a limiting case of LHS and also as an improvement over it. An experiment and a set of empirical results illustrating the relationship between DS and LHS are also presented.

Supply chain analysis: spreadsheet or simulation?

In the last few decades, a lot of company effort has been spent in the optimization of internal efficiency, aiming at cost reduction and competitiveness. Especially over the last decade, there has been a consensus that not only the company, but the whole supply chain in which it fits, is responsible for the success or failure of any business. Therefore, supply chain analysis tools and methodologies have become more and more important. From all tools, spreadsheets are by far the most widely used technique for scenario analysis. Other techniques such as optimization, simulation or both (simulation-optimization) are alternatives for in-depth analysis. While spreadsheet-based analysis is mainly a static-deterministic approach, simulation is a dynamic-stochastic tool. The purpose of this paper is to compare spreadsheet-based and simulation-based tools showing the impacts of using these two different approaches on the analysis of a real (yet simplified) supply chain case study.

An empirical evaluation of sampling methods in risk analysis simulation: quasi-Monte Carlo, descriptive sampling, and latin hypercube sampling

This paper compares the performance, in terms of convergence rates and precision of the estimates, for six Monte Carlo simulation sampling methods: quasi-Monte Carlo using Halton, Sobol, and Faure numeric sequences; descriptive sampling, based on the use of deterministic sets and Latin hypercube sampling, based on stratified numerical sets. Those methods are compared to the classical Monte Carlo. The comparison was made for two basic risky applications: the first one evaluates the risk in a decision making process when launching a new product; the second evaluates the risk of accomplishing an expected rate of return in a correlated stock portfolio. Descriptive sampling and Latin hypercube sampling have shown the best aggregate results.

Cooperação entre redes neurais artificiais e técnicas 'clássicas' para previsão de demanda de uma série de vendas de cerveja na Austrália

The main goal of this article is to evaluate the complementation between forecasting techniques, in particular Neural Networks. The environment of the application is a foreign industrial sector, sales of beer in Australia. Most of the forecasting models act in a separate way, that is, treating problems through different views that exclude one another. The suggestion of this work is to use the methods in a cooperative way, looking forward to achieve better results. The article begins with some considerations on Time Series Studies, followed by a brief characterization of Neural Networks and the other methods used on this work. Continuing, the case is presented and discussed and finishing, conclusions and a list of references.

A farewell to the use of antithetic variates in Monte Carlo simulation

Antithetic variates (AV) is one the oldest and most popular variance reduction techniques (VRTs), commonly using complementary random numbers. The AV variance reduction is generally justified by the negative correlation it produces in paired simulation estimates. A new and simpler interpretation of the AV role is presented, showing AV as solely a procedure for input sample means compensation, without any further contribution from the complementary idea. The proposed interpretation is based on the descriptive sampling framework, viewing input samples as composed of a set of values and their sequencing. Simulation experiments and third-party results give support to this interpretation. However, when newer simulation sampling methods, like Latin Hypercube Sampling, Descriptive Sampling, Moment Matching and Quasi-Monte Carlo are adopted, all of them based on a controlled selection of the input sample values, AV turns irrelevant. Other VRTs are also affected by the ideas presented here.

Out-of-the-money monte carlo simulation option pricing: the joint use of importance sampling and descriptive sampling

As in any Monte Carlo application, simulation option valuation produces imprecise estimates. In such an application, descriptive sampling (DS) has proven to be a powerful variance reduction technique. However, this performance deteriorates as the probability of exercising an option decreases. In the case of out-of-the-money options, the solution is to use importance sampling (IS). Following this track, the joint use of IS and DS is deserving of attention. Here, we evaluate and compare the benefits of using standard IS method with the joint use of IS and DS. We also investigate the influence of the problem dimensionality in the variance reduction achieved. Although the combination IS+DS showed gains over the standard IS implementation, the benefits in the case of out-of-the-money options were mainly due to the IS effect. On the other hand, the problem dimensionality did not affect the gains. Possible reasons for such results are discussed.

Proposta para a gestão de estoques de novos produtos: solução do modelo (Q,r) para a distribuição uniforme da demanda e do lead-time de suprimento

The premises commonly adopted in inventory management models - adherence of lead-time demand to Normal Distribution, known average and standard deviations, and discrete lead time - are often unrealistic and can lead to considerable distortions in new product inventories, particularly insofar as total costs and service level indicators are concerned. Considering that companies stock new products while simultaneously learning about the characteristics of lead-time demand distribution, this paper proposes the solution to the (Q,r) - order quantity and reorder point - inventory model for uniform demand and lead-time. The premise of Uniform Distribution is defined by two parameters that are more intuitive than mean and standard deviation - maximum and minimum - and it can also be applied when a result shows the same probability of occurring. Therefore, its adoption may be the first practical approach for new product inventory management.

Soccer championship analysis using Monte Carlo simulation

Sports had always fascinated humanity. In this context, soccer was taken as a study source. The objective of this paper is to formulate a simulation model to generate estimators for necessary scores to achieve certain places at the final classification ranking of the Brazilian National Soccer Championship. The main data used are the rules of the championship, the number of competitors and the probability that a match ends up in a draw.

Top-down or bottom-up forecasting?

The operations literature continues on inconclusive as to the most appropriate sales forecasting approach (Top-Down or Bottom-up) for the determination of safety inventory levels. This paper presents the analytical results for the variance of the sales forecasting errors during the lead-time in both approaches. The forecasting method used was the Simple Exponential Smoothing and the results led to the identification of two supplementary impacts upon the forecasting error variance, and consequently, upon safety inventory levels: the Portfolio Effect and the Anchoring Effect. The first depends upon the correlation coefficient of demand between two individual items and the latter, depends upon the smoothing constant and upon the participation of the individual item in total sales. It is also analysed under which conditions these variables would favour one forecasting approach instead of the other.

A simulation model to validate and evaluate the adequacy of an analytical expression for proper safety stock sizing

The purpose of this paper is to validate and test the adequacy of an analytical expression to calculate proper safety stock levels using simulation techniques. The model refers to a periodic review system and a lot-4-lot replenishment policy, with randomness in forecast errors and in order fulfillment. The simulation model is formulated in a spreadsheet environment using MS Excel/sup /spl reg// and @Risk/sup /spl reg//. The percentage of periods without stockout is computed and compared to the theoretical value expected by the assumptions inherent to the analytical expression.