Statnote 23: Non-parametric analysis of variance (ANOVA)

To carry out an analysis of variance, several assumptions are made about the nature of the experimental data which have to be at least approximately true for the tests to be valid. One of the most important of these assumptions is that a measured quantity must be a parametric variable, i.e., a member of a normally distributed population. If the data are not normally distributed, then one method of approach is to transform the data to a different scale so that the new variable is more likely to be normally distributed. An alternative method, however, is to use a non-parametric analysis of variance. There are a limited number of such tests available but two useful tests are described in this Statnote, viz., the Kruskal-Wallis test and Friedmann’s analysis of variance.

Non-normally distributed data and non-parametric statistics

Different types of numerical data can be collected in a scientific investigation and the choice of statistical analysis will often depend on the distribution of the data. A basic distinction between variables is whether they are ‘parametric’ or ‘non-parametric’. When a variable is parametric, the data come from a symmetrically shaped distribution known as the ‘Gaussian’ or ‘normal distribution’ whereas non-parametric variables may have a distribution which deviates markedly in shape from normal. This article describes several aspects of the problem of non-normality including: (1) how to test for two common types of deviation from a normal distribution, viz., ‘skew’ and ‘kurtosis’, (2) how to fit the normal distribution to a sample of data, (3) the transformation of non-normally distributed data and scores, and (4) commonly used ‘non-parametric’ statistics which can be used in a variety of circumstances.

Has DRG payment influenced the technical efficiency and productivity of diagnostic technologies in Portuguese public hospitals? An empirical analysis using parametric and non‐parametric methods

The use of Diagnosis Related Groups (DRG) as a mechanism for hospital financing is a currently debated topic in Portugal. The DRG system was scheduled to be initiated by the Health Ministry of Portugal on January 1, 1990 as an instrument for the allocation of public hospital budgets funded by the National Health Service (NHS), and as a method of payment for other third party payers (e.g., Public Employees (ADSE), private insurers, etc.). Based on experience from other countries such as the United States, it was expected that implementation of this system would result in more efficient hospital resource utilisation and a more equitable distribution of hospital budgets. However, in order to minimise the potentially adverse financial impact on hospitals, the Portuguese Health Ministry decided to gradually phase in the use of the DRG system for budget allocation by using blended hospitalspecific and national DRG casemix rates. Since implementation in 1990, the percentage of each hospitals budget based on hospital specific costs was to decrease, while the percentage based on DRG casemix was to increase. This was scheduled to continue until 1995 when the plan called for allocating yearly budgets on a 50% national and 50% hospitalspecific cost basis. While all other nonNHS third party payers are currently paying based on DRGs, the adoption of DRG casemix as a National Health Service budget setting tool has been slower than anticipated. There is now some argument in both the political and academic communities as to the appropriateness of DRGs as a budget setting criterion as well as to their impact on hospital efficiency in Portugal. This paper uses a twostage procedure to assess the impact of actual DRG payment on the productivity (through its components, i.e., technological change and technical efficiency change) of diagnostic technology in Portuguese hospitals during the years 1992–1994, using both parametric and nonparametric frontier models. We find evidence that the DRG payment system does appear to have had a positive impact on productivity and technical efficiency of some commonly employed diagnostic technologies in Portugal during this time span.

Assessing pupil and school performance by non-parametric and parametric techniques

Often observations are nested within other units. This is particularly the case in the educational sector where school performance in terms of value added is the result of school contribution as well as pupil academic ability and other features relating to the pupil. Traditionally, the literature uses parametric (i.e. it assumes a priori a particular function on the production process) Multi-Level Models to estimate the performance of nested entities. This paper discusses the use of the non-parametric (i.e. without a priori assumptions on the production process) Free Disposal Hull model as an alternative approach. While taking into account contextual characteristics as well as atypical observations, we show how to decompose non-parametrically the overall inefficiency of a pupil into a unit specific and a higher level (i.e. a school) component. By a sample of entry and exit attainments of 3017 girls in British ordinary single sex schools, we test the robustness of the non-parametric and parametric estimates. We find that the two methods agree in the relative measures of the scope for potential attainment improvement. Further, the two methods agree on the variation in pupil attainment and the proportion attributable to pupil and school level.

When do conservation planning methods deliver? Quantifying the consequences of uncertainty

The rapid global loss of biodiversity has led to a proliferation of systematic conservation planning methods. In spite of their utility and mathematical sophistication, these methods only provide approximate solutions to real-world problems where there is uncertainty and temporal change. The consequences of errors in these solutions are seldom characterized or addressed. We propose a conceptual structure for exploring the consequences of input uncertainty and oversimpli?ed approximations to real-world processes for any conservation planning tool or strategy. We then present a computational framework based on this structure to quantitatively model species representation and persistence outcomes across a range of uncertainties. These include factors such as land costs, landscape structure, species composition and distribution, and temporal changes in habitat. We demonstrate the utility of the framework using several reserve selection methods including simple rules of thumb and more sophisticated tools such as Marxan and Zonation. We present new results showing how outcomes can be strongly affected by variation in problem characteristics that are seldom compared across multiple studies. These characteristics include number of species prioritized, distribution of species richness and rarity, and uncertainties in the amount and quality of habitat patches. We also demonstrate how the framework allows comparisons between conservation planning strategies and their response to error under a range of conditions. Using the approach presented here will improve conservation outcomes and resource allocation by making it easier to predict and quantify the consequences of many different uncertainties and assumptions simultaneously. Our results show that without more rigorously generalizable results, it is very dif?cult to predict the amount of error in any conservation plan. These results imply the need for standard practice to include evaluating the effects of multiple real-world complications on the behavior of any conservation planning method.

Statnote 4: what if the data are not normal?

When testing the difference between two groups, if previous data indicate non-normality, then either transform the data if they comprise percentages, integers or scores or use a non-parametric test. If there is uncertainty whether the data are normally distributed, then deviations from normality are likely to be small if the data are measurements to three significant figures. Unless there is clear evidence that the distribution is non-normal, it is more efficient to use the conventional t-tests. It is poor statistical practice to carry out both the parametric and non-parametric tests on a set of data and then choose the result that is most convenient to the investigator!

Statnote 15: the non-parametric correlation coefficients

If in a correlation test, one or both variables are small whole numbers, scores based on a limited scale, or percentages, a non-parametric correlation coefficient should be considered as an alternative to Pearson’s ‘r’. Kendall’s t and Spearman’s rs are similar tests but the former should be considered if the analysis is to be extended to include partial correlations. If the data contain many tied values, then gamma should be considered as a suitable test.

Pattern generation by dissipative parametric instability

Nonlinear instabilities are responsible for spontaneous pattern formation in a vast number of natural and engineered systems, ranging from biology to galaxy buildup. We propose a new instability mechanism leading to pattern formation in spatially extended nonlinear systems, which is based on a periodic antiphase modulation of spectrally dependent losses arranged in a zigzag way: an effective filtering is imposed at symmetrically located wave numbers k and -k in alternating order. The properties of the dissipative parametric instability differ from the features of both key classical concepts of modulation instabilities, i.e., the Benjamin-Feir instability and the Faraday instabiltyity. We demonstrate how the dissipative parametric instability can lead to the formation of stable patterns in one- and two-dimensional systems. The proposed instability mechanism is generic and can naturally occur or can be implemented in various physical systems.

Neural Network Modelling with Input Uncertainty: Theory and Application

It is generally assumed when using Bayesian inference methods for neural networks that the input data contains no noise. For real-world (errors in variable) problems this is clearly an unsafe assumption. This paper presents a Bayesian neural network framework which accounts for input noise provided that a model of the noise process exists. In the limit where the noise process is small and symmetric it is shown, using the Laplace approximation, that this method adds an extra term to the usual Bayesian error bar which depends on the variance of the input noise process. Further, by treating the true (noiseless) input as a hidden variable, and sampling this jointly with the network’s weights, using a Markov chain Monte Carlo method, it is demonstrated that it is possible to infer the regression over the noiseless input. This leads to the possibility of training an accurate model of a system using less accurate, or more uncertain, data. This is demonstrated on both the, synthetic, noisy sine wave problem and a real problem of inferring the forward model for a satellite radar backscatter system used to predict sea surface wind vectors.

Managing uncertainty in complex stochastic models: design and emulation of a Rabies model

In this paper we present a novel method for emulating a stochastic, or random output, computer model and show its application to a complex rabies model. The method is evaluated both in terms of accuracy and computational efficiency on synthetic data and the rabies model. We address the issue of experimental design and provide empirical evidence on the effectiveness of utilizing replicate model evaluations compared to a space-filling design. We employ the Mahalanobis error measure to validate the heteroscedastic Gaussian process based emulator predictions for both the mean and (co)variance. The emulator allows efficient screening to identify important model inputs and better understanding of the complex behaviour of the rabies model.

An Exploration of parametric software cost estimating models for Jackson Systems Development

Most parametric software cost estimation models used today evolved in the late 70's and early 80's. At that time, the dominant software development techniques being used were the early 'structured methods'. Since then, several new systems development paradigms and methods have emerged, one being Jackson Systems Development (JSD). As current cost estimating methods do not take account of these developments, their non-universality means they cannot provide adequate estimates of effort and hence cost. In order to address these shortcomings two new estimation methods have been developed for JSD projects. One of these methods JSD-FPA, is a top-down estimating method, based on the existing MKII function point method. The other method, JSD-COCOMO, is a sizing technique which sizes a project, in terms of lines of code, from the process structure diagrams and thus provides an input to the traditional COCOMO method.The JSD-FPA method allows JSD projects in both the real-time and scientific application areas to be costed, as well as the commercial information systems applications to which FPA is usually applied. The method is based upon a three-dimensional view of a system specification as opposed to the largely data-oriented view traditionally used by FPA. The method uses counts of various attributes of a JSD specification to develop a metric which provides an indication of the size of the system to be developed. This size metric is then transformed into an estimate of effort by calculating past project productivity and utilising this figure to predict the effort and hence cost of a future project. The effort estimates produced were validated by comparing them against the effort figures for six actual projects.The JSD-COCOMO method uses counts of the levels in a process structure chart as the input to an empirically derived model which transforms them into an estimate of delivered source code instructions.

Aspects of the design and control of manufacturing systems subject to demand uncertainty

The recent explosive growth in advanced manufacturing technology (AMT) and continued development of sophisticated information technologies (IT) is expected to have a profound effect on the way we design and operate manufacturing businesses. Furthermore, the escalating capital requirements associated with these developments have significantly increased the level of risk associated with initial design, ongoing development and operation. This dissertation has examined the integration of two key sub-elements of the Computer Integrated Manufacturing (CIM) system, namely the manufacturing facility and the production control system. This research has concentrated on the interactions between production control (MRP) and an AMT based production facility. The disappointing performance of such systems has been discussed in the context of a number of potential technological and performance incompatibilities between these two elements. It was argued that the design and selection of operating policies for both is the key to successful integration. Furthermore, policy decisions are shown to play an important role in matching the performance of the total system to the demands of the marketplace. It is demonstrated that a holistic approach to policy design must be adopted if successful integration is to be achieved. It is shown that the complexity of the issues resulting from such an approach required the formulation of a structured design methodology. Such a methodology was subsequently developed and discussed. This combined a first principles approach to the behaviour of system elements with the specification of a detailed holistic model for use in the policy design environment. The methodology aimed to make full use of the `low inertia' characteristics of AMT, whilst adopting a JIT configuration of MRP and re-coupling the total system to the market demands. This dissertation discussed the application of the methodology to an industrial case study and the subsequent design of operational policies. Consequently a novel approach to production control resulted. A central feature of which was a move toward reduced manual intervention in the MRP processing and scheduling logic with increased human involvement and motivation in the management of work-flow on the shopfloor. Experimental results indicated that significant performance advantages would result from the adoption of the recommended policy set.

Longitudinal parametric oscillation in photorefractive sillenites: comparison between theory and experiment

By using an alternative setup for photorefractive parametric oscillation in which wave mixing between the recording beams is avoided it has become possible to make more detailed comparisons with the space-charge wave theory. In the present paper we compare the experimental features of longitudinal parametric oscillation observed in a crystal of Bi12SiO20 with the theoretical predictions.

Influence of beam-coupling on photorefractive parametric oscillation in a dc-field-biased Bi12SiO20 crystal

The influence of beam coupling on photorefractive parametric oscillation generated in a Bi12SiO20 crystal is investigated experimentally by comparing two configurations with and without the presence of beam coupling. It is shown that beam coupling has a great influence; for example, the transversal split of the K/2 subharmonic grating is seen only in the beam-coupling geometry. A case that resembles K/4 subharmonic generation can, however, still be found in the absence of beam coupling.