Mezbahur Rahman

Home page usability and credibility: A comparison of the fastest growing companies to the Fortune 30 and the implications to IT governance

Purpose – The purpose of the research was to compare web site designs used by the fastest growing companies in the USA to the largest companies in the USA and to benchmark those designs against best practices as defined by a leading consultant in the industry.Design/methodology/approach – This approach surveyed the web site designs by each group of companies against a set of best practices and developed summary data about the observations.Findings – The largest companies in the USA used designs that are more consistent with the best practices as defined by a leading consultant in the industry.Research limitations/implications – The fastest growing companies did not use best practices in their web site design and still maintained very rapid growth as evidenced by their national rankings. A poorly executed design was less of a detractor to the companys success than anticipated. Web site design may have a role (positive or negative) in the companys growth rate but further research is required to understand...

Estimation in two-parameter exponential distributions

Exponential distributions are used extensively in the field of life-testing. Estimation of parameters is revisited in two-parameter exponential distributions. A comparison study between the maximum likelihood method, the unbiased estimates which are linear functions of the maximum likelihood method, the method of product spacings, and the method of quantile estimates are presented. Finally, a simulation study is given to demonstrate the small sample properties

Estimating the box-cox transformation via shapiro-wilk W Statistic

The Box-Cox transformation is a well known family of power transformations to bring a set of data into agreement with the normality assumption of the residuals and hence the response variable of a postulated model in regression analysis. This paper proposes a new method for estimating the Box-Cox transformation using maximization of the Shapiro-Wilk W statistic which forces the data to get closer to normal as much as possible. A comparative study of the proposed procedure with the normal based likelihood procedure and the artificial regression model procedure also presented.

A note on combining parametric and non-parametric regression

A combination of a parametric estimate and a nonparametric estimate of a model for a regression function is considered. The optimal linear combination is estimated from the data by the least squares estimate of the combining coefficient. The estimate so obtained is compared with the one proposed by Wooldridge (1992) and Burman and Chaud-huri,the latter being based on Stein (1956). A test procedure for deciding about the parametric specification is also studied.

Quantiles for finite mixtures of normal distributions

Quantiles for finite mixtures of normal distributions are computed. The difference between a linear combination of independent normal random variables and a linear combination of independent normal densities is emphasized.

A note on estimating parameters in two-parameter Pareto distributions

Pareto distributions are used extensively in modelling income distributions. Estimation of parameters is revisited in two-parameter Pareto distributions. The method of quantile estimates using the elemental estimates and the method of product spacings are applied to the two-parameter Pareto distributions. A comparative study between the maximum likelihood method, the unbiased estimates which are functions of the maximum likelihood method, the minimum mean squared error method, the method of moments, the method of quantile estimation, the method of quantile estimation using the elemental estimates and the method of product spacings is presented.

On estimation of parameters of the exponential power family of distributions

Consider the three-parameter exponential power distribution with location parameter μ, scale parameter σ2 and shape (power) parameter β. This is a general symmetric family of distributions with normal, double exponential and rectangular as special cases. Such distributions are used in Bayesian statistics as a wider choice of symmetric parent distribution and in classical statistics in determination of lack of normality. This note obtains simultaneous estimates of μ, σ2 and β by method of moments and method of maximum likelihood. It also studies the behavior of estimates of β through Monte Carlo simulation when values of μ and σ2 are set equal to zero and unity respectively.

A note on logistic regression

Abstract Maximum likelihood and discriminant analysis in estimating logistic regression parameters are revisited. Minimization of mis-prediction is considered as the criteria of the goodness of the model. It is shown that the parameter estimates are more accurate in terms of higher correct prediction rate in case of discriminant analysis compared to the maximum likelihood procedure for smaller samples. For larger samples, even-though the parameter estimates in the maximum likelihood procedure are more theoretically sound but the discriminant analysis procedure is as good in terms of percentages of correct prediction. A bilingual education data is analysed in the scope of the estimation procedure mentioned in this paper.

Attribute relative importance computation in conjoint analysis

Abstract In conjoint analysis, computation of relative importance for different attributes play vital role in deciding about elimination or inclusion of any attribute. In the literature, relative ranges of factor effects at different levels are considered as the measure for relative importance weights. We consider the well known reduced sum of squares procedure to decide about attribute importance weights. Such constructed weights have better interpretation in terms of portion of total variability explained by the attribute in the model.

On testimation of a probability density: the normal case

A two-stage shrinkage testimation procedure is discussed in the problem of density estimation. A first stage sample is used to test the goodness of fit of the data compared with a pre-assumed hypothesis about the form of the density function. A combination of parametric and nonparametric estimates is taken as the estimate of the density if the data agree with the hypothesis. Otherwise, a second sample is taken and a purely nonparametric estimate is obtained using the combined sample. The procedure protects against model misspecification and results in taking smaller samples when inference can be based only on the first stage sample.