Determining the parametric structure of models.

In this paper we develop a comprehensive approach to determining the parametric structure of models. This involves considering whether a model is parameter redundant or not and investigating model identifiability. The approach adopted makes use of exhaustive summaries, quantities that uniquely define the model. We review and generalise previous work on evaluating the symbolic rank of an appropriate derivative matrix to detect parameter redundancy, and then develop further tools for use within this framework, based on a matrix decomposition. Complex models, where the symbolic rank is difficult to calculate, may be simplified structurally using reparameterisation and by finding a reduced-form exhaustive summary. The approach of the paper is illustrated using examples from ecology, compartment modelling and Bayes networks. This work is topical as models in the biosciences and elsewhere are becoming increasingly complex.

Gaussian process functional regression modeling for batch data.

A Gaussian process functional regression model is proposed for the analysis of batch data. Covariance structure and mean structure are considered simultaneously, with the covariance structure modeled by a Gaussian process regression model and the mean structure modeled by a functional regression model. The model allows the inclusion of covariates in both the covariance structure and the mean structure. It models the nonlinear relationship between a functional output variable and a set of functional and nonfunctional covariates. Several applications and simulation studies are reported and show that the method provides very good results for curve fitting and prediction.

Beyond the binary Boltzmann machine.

The definition of the usual Boltzmann machine is extended to allow for neurons with polytomous (multicategory) rather than simply binary responses. Updating rules are defined along with the associated stationary distributions, and an alternating minimization method is described for the purposes of training. Emphasis is placed on the relevance of statistical ideas, including polytomous logistic regression, the iterative proportional fitting procedure and the EM algorithm.

On some Bayesian/regularization methods for image restoration.

Methods are reviewed for choosing regularized restorations in image processing. In particular, a method developed by Galatsanos and Katsaggelos (see ibid., vol.1, p.322-336, 1992) is given a Bayesian interpretation and is compared with other Bayesian and non-Bayesian alternatives. A small illustrative example is provided and a complement is provided for the discussion of noise variance estimation of Galatsanos et al.

Discriminant analysis of macrocytic red cells.

Laboratory classification of red cell disorders uses the red cell indices (MCV, MCH, MCHC, RDW) and information gleaned from microscopic evaluation of a blood film. Additional red cell information is now available using the H series of automated blood cell analysers (Ames Technicon Division of Bayer Diagnostics). This study involved the development of a discriminant rule which would differentiate between three causes of macrocytosis (vitamin B12/folate deficiency, alcohol excess/liver disease and a reticulocytosis) using the information available on Research Screen 1 and Report Screen 3 of the H*1 instrument (Report Screen 3 is a graphical display of the trimmed scattergram of red cell volume and red cell haemoglobin concentration and Research Screen 1 displays the associated numerical data). Three methods of analysis were assessed to define a suitable discriminant rule. The percentages of patients correctly classified by the three methods were: 92.1%, 82.0% and 89.2% for Methods 1, 2 and 3 respectively. Method 1 involved the application of quadratic discrimination to transformed variables and produced the best results. Although complex, it could easily be applied using the microprocessor capability of the average multiparameter haematology analyser.

The quality and relevance of peripheral neuropathy data on a diabetic clinical information system.

Routinely collected peripheral neuropathy data entered on a diabetic clinical information system since 1979 have been audited for completeness, consistency, accuracy (inter-observer variation), validity by comparison with biothesiometry, and relevance by life table analysis for foot ulceration. Peripheral neuropathy was defined by a neuropathy disability score > or = 4. The data were 98% complete. Forty-nine of 3405 (1.4%) had inconsistent records. Agreement between observers for clinical examination was significant (p < 0.05) for aggregate neuropathy score and its individual components except the knee jerk: Kappa score for observer variation for neuropathy score 0.56 (95% confidence interval 0.36-0.76). There was good agreement between neuropathy defined as aggregate score > or = 4, and as combined vibration perception thresholds for both feet > 60 V: Kappa statistic 0.62 (95% confidence interval 0.44-0.80). The chance of developing a foot problem in 3 years increased from 3% for patients with a score of zero to 45% for people with a score of between 9 and 12. We conclude that the calculation of a clinical neuropathy score is a simple, valid and relevant method for diabetes care both in hospital and the community. When combined with palpation of peripheral pulses most patients at risk of foot ulceration can be identified allowing targeting of preventive chiropody and orthotic resources.

Ophthalmic consequences of mid-facial trauma.

Fractures of the mid-face are commonly accompanied by injury to the visual system. Three hundred and sixty three patients who had sustained mid-facial fractures were assessed prospectively for evidence of such injury. The data have been used to devise a scoring system for the maxillofacial surgeon in order to help identify those patients at risk of eye injury. The scoring system has been evaluated on a further cohort of 100 patients with a similar spectrum of injury and the sensitivity and specificity of the system have been determined. The results of these studies have been reported in the maxillofacial literature. This paper reviews the data and results obtained. In summary, impaired visual acuity with a comminuted or out blow fracture, a motility abnormality, or facial fracture combined with head injury, sufficient to cause both retrograde and post-traumatic amnesia, emerged as major risk factors which are indicative of an adverse ophthalmic outcome deemed to warrant referral. The scoring system which was developed from this data was found to have a sensitivity of 94.4% and a specificity of 89% for the detection of patients thought to merit ophthalmic assessment. Failure to assess central visual function as objectively as practicable in patients who have sustained mid-facial fractures may lead to potentially treatable ophthalmic pathology not being identified.

Midfacial fractures and the eye: the development of a system for detecting patients at risk of eye injury.

Maxillofacial trauma is often complicated by injury to the eye. Such injuries may be difficult to detect and may therefore be missed. Detailed ophthalmic examinations were carried out prospectively on 363 patients who had sustained midfacial fractures. Fifty four parameters comprising maxillofacial, radiological and ophthalmic data were recorded and coded for each patient. All encoded data were divided into predictors (the data potentially available to the maxillofacial surgeon) and outcome (the data potentially available to the ophthalmologist). Statistical methods of regression, and the analysis of contingency tables, led to the identification of the principal predictors indicative of underlying ophthalmic injury and thence to a scoring system which predicts the severity of such injuries. Impaired visual acuity is the principal predictor and when employed alone gives a sensitivity value of 80%. Pure blow-out fracture or comminuted facial fracture, double vision and amnesia emerged as additional factors which yielded an efficient scoring system with a sensitivity of 89% and specificity of 90% for the population upon which it was based. A score sheet is provided in the paper. These predictors can be remembered from the acronym Blow-out fracture, Acuity, Diplopia, Amnesia, Comminuted Trauma. As many such injuries result from a BAD ACT, it is easily remembered. This scoring system requires to be tested upon a new population of individuals in order to determine its efficacy.

Midfacial fractures and the eye: the development of a system for detecting patients at risk of eye injury--a prospective evaluation.

Midfacial trauma is often complicated by ocular disorder. A scoring system has been devised to help the maxillofacial surgeon identify patients who warrant referral to an ophthalmologist. A prospective pilot study was carried out on 100 patients with midfacial fractures to evaluate the effectiveness of this system in clinical practice. The sensitivity value was 94.4% and the specificity value was 89%. Only 1 patient, who clinically warranted referral to an ophthalmologist, was missed by the system whilst 9 others were incorrectly classified as warranting referral. The results of this evaluation demonstrate the competence of the system.

An optimal design problem in rhythmometry.

A trigonometric regression model is assumed for a problem involving circadian rhythm exhibited by peak expiratory flow. Experimental designs are sought with a view to estimating a particular nonlinear function of the parameters. Both optimal and nonoptimal, but more practicable, designs are derived and their relative efficiencies are established.

On the Iterative Image Space Reconstruction Algorthm for ECT.

The ISRA of [1] is shown to be an iterative algorithm that aims to converge to the least-squares estimates of emission densities. Convergence is established in the case where a unique least-squares estimate exists that is, elementwise, strictly positive. It is pointed out that, in terms of asymptotic theory, the resulting estimators are inferior to the maximum likelihood estimators, for which the EM algorithm is a computational procedure. Potential difficulties with the behavior of the ISRA are illustrated using very simple examples.

Injection site lesions of footrot vaccines in sheep.

Three trials were conducted to assess the size and frequency of lesions induced by an oil-based and an alum-precipitated aqueous B. nodosus commercial footrot vaccine at two injection sites (neck and brisket) with or without prior skin swabbing with methanol. Animals used were previously unvaccinated and clean and dry at time of vaccination. Trial 1 contained 57 2-tooth Booroola Merino x Coopworth ewes, trial 2 contained 60 mixed-age Corriedale ewes, and trial 3 contained 39 rams of mixed age and sex. Firm nodules or plaques with indurated edges and with a maximum dimension of 1cm or more appeared soon after vaccination and were still very common (50-100% of sheep vaccinated) 5-6 weeks after injection. A variable proportion exhibited a liquefied central area of necrosis and later discharged necrotic material. The oil-based vaccine produced significantly larger lesions than the alum-precipitated aqueous vaccine. Lesion size at the brisket site was significantly larger than the neck site in ewes but not in rams. Skin preparation with methanol had no effect on lesion frequency or size.

On constrained balance randomization for clinical trials.

A method is proposed for calculating the probabilities of assignment of a patient to treatments; it involves minimizing a quadratic criterion subject to a balance constraint. The optimal probabilities are very easy to compute. Numerical illustration is given and comparisons are drawn with the entropy-based methods of Klotz (1978, Biometrics 34, 283-287).

Treatment of cold urticaria.

In a single patient with severe essential acquired cold urticaria, the following drug regimes were assessed: diphenhydramine, cyproheptadine, hydroxyzine, terbutaline, cimetidine, hydroxyzine plus cimetidine, and hydroxyzine plus terbutaline. Each regime showed favourable results, but a significance level of P = 0.01 was reached only for (a) hydroxyzine plus cimetidine in the suppression of erythema, and (b) the histamine H1 antagonists in the suppression of wealing.

Relation between prognosis and the blood pressure before and during treatment of hypertensive patients.

1. A study was conducted amongst 1247 treated hypertensive patients to determine the predictive power of untreated baseline and achieved treated blood pressures in the development of the complications of hypertension. In addition the relative importance of systolic and diastolic pressures was calculated. 2. Statistical analysis was done by calculating univariate differences in blood pressure between cases with and without complications. The higher the univariate distance, the greater the predictive power. 3. Blood pressures achieved during treatment were more important than baseline pressures for predicting stroke in both men and women, confirming the benefits of antihypertensive therapy in preventing strokes. 4. There was some evidence of prevention of myocardial infarction in men and of angina in women as a result of therapy. 5. There was no evidence to suggest that any one group of drugs, including beta-adrenoreceptor-blocking drugs and thiazides, conferred any extra benefit in preventing coronary heart disease. 6. The systolic blood pressures achieved during treatment predicted stroke better than diastolic pressure, but no consistent trends were found for coronary heart disease.

Influence of HLA matching and blood transfusion on renal allograft survival.

One hundred and seven consecutive cadaver kidney transplants have been followed for up to 6 years. The beneficial effect of HLA matching, shown in previous studies, has been confirmed. The 2-year failure rate from rejection was 29% for grafts with less than two incompatibilities, in comparison with a figure of 52% where there were two or more incompatibilities. In contrast to some reports, the presence of HLA antibodies did not have an adverse effect on the survival of first grafts. Patients not transfused prior to transplantation had a much higher 1-year graft failure rate (72%) than those given either frozen-thawed red cells (29%) or whole blood (23%). This apparently beneficial effect of blood transfusion was no greater in patients transfused with more than five units compared with those given less than five units. We believe that blood transfusion has an important influence on the outcome of renal transplantation.