Kenneth G. Russell

Designs for generalized linear models with several variables and model uncertainty

Standard factorial designs sometimes may be inadequate for experiments that aim to estimate a generalized linear model, for example, for describing a binary response in terms of several variables. A method is proposed for finding exact designs for such experiments that uses a criterion allowing for uncertainty in the link function, the linear predictor, or the model parameters, together with a design search. Designs are assessed and compared by simulation of the distribution of efficiencies relative to locally optimal designs over a space of possible models. Exact designs are investigated for two applications, and their advantages over factorial and central composite designs are demonstrated.

Validation of an Australian electronic food frequency questionnaire to measure polyunsaturated fatty acid intake

OBJECTIVE To develop and validate a simple non-invasive method that estimates the intakes of omega-3 and omega-6 polyunsaturated fatty acids (PUFA) in a healthy adult population. METHODS A new electronic PUFA food frequency questionnaire (FFQ) was validated by comparison with a 3-d weighed food record and blood biomarkers (erythrocytes and plasma) using the method of triads model and tested for reproducibility. Healthy subjects were recruited from the local Illawarra Region, New South Wales, Australia. RESULTS The PUFA FFQ adequately estimated intakes for eicosapentaenoic acid, docosahexaenoic acid, total long chain omega-3 PUFA, linoleic acid, total omega-6 PUFA, and total PUFA, which were comparable with results from the 3-d food record. Eicosapentaenoic acid, docosahexaenoic acid, and total long chain omega-3 showed high validity coefficients for erythrocytes (and plasma) 0.92 (0.87), 0.69 (0.64), and 0.78 (0.73) (P < 0.05), respectively. Spearmans rank correlation coefficients ranged from 0.48 to 0.76 when the PUFA FFQ was tested for reproducibility (P < 0.05). CONCLUSION The electronic PUFA questionnaire was found to be reproducible and is a valid tool to assess PUFA intakes in a healthy adult population.

Using interviews and peer pairs to better understand how school environments affect young children’s playground physical activity levels: a qualitative study

School break times provide a daily opportunity for children to be active; however, research indicates this time is underutilized. Reasons for low childrens playground activity levels have primarily focused on physical barriers. This research aimed to contribute to physical environmental findings affecting childrens playground physical activity levels by identifying additional variables through the interview process. Thirteen public schools were included in the sample (total 2946 children). Physical activity and environmental data were collected over 3 days. Environmental variables were manually assessed at each of the 13 schools. Observational data were used to determine which three schools were the most and least active. The principal, three teachers and 20 students in Grades 4-6 from these six schools (four lower and two average socioeconomic status) were invited to participate in the interview process. Student interviews involved the paired interview technique. The main themes generated from the school interviews included the effect of non-fixed equipment (including balls), playground markings, playground aesthetics, activity preference, clothing, the amount of break time available for play, teacher playground involvement, gender, bullying, school policies, student confidence in break-time activity and fundamental movement skills. The effect of bullying on playground physical activity levels was concerning.

Relative bias in diet history measurements: a quality control technique for dietary intervention trials.

OBJECTIVE Investigation of relative bias in diet history measurement during dietary intervention trials. DESIGN Retrospective analysis of human dietary data from two randomised controlled trials examining modified fat diets in the prevention and treatment of type II diabetes mellitus. SETTING Wollongong, Australia. SUBJECTS Thirty-five overweight, otherwise healthy subjects in trial 1 and 56 subjects with diabetes in trial 2. INTERVENTIONS Diet history interviews and three-day weighed food records administered at one-month intervals in trial 1 and three-month intervals in trial 2. RESULTS In a cross-sectional bias analysis, graphs of the association between bias and mean dietary intake showed that bias decreased in higher carbohydrate consumers in trial 1 (r = -0.344, P < 0.05). No other significant associations were found. In a longitudinal analysis, bias did not change over time in either trial. There were no significant differences in bias magnitudes between the trials, with the exception of monounsaturated fat measurement where bias was significantly greater and more positive in trial 2, indicating overestimation of monounsaturated fat intake with the diet history. Subjects in control and intervention groups underestimated energy, fat, saturated fat and alcohol intakes with the diet history in both trials. Overweight and obese individuals appeared to make the greatest contribution to the overall underestimation of saturated fat intake by the diet history regardless of whether they were in the control or intervention group and whether they were healthy or had diabetes. CONCLUSION Bias in diet history measurement appears to be macronutrient-specific, with energy, fat and saturated fat consistently underreported in the interview by subjects with and without diabetes and in both intervention and control groups in a dietary intervention trial. Relative bias analysis appears to be an informative tool in quality control for dietary intervention trials when biochemical markers are unavailable.

Design considerations for small experiments and simple logistic regression

Inference for a generalized linear model is generally performed using asymptotic approximations for the bias and the covariance matrix of the parameter estimators. For small experiments, these approximations can be poor and result in estimators with considerable bias. We investigate the properties of designs for small experiments when the response is described by a simple logistic regression model and parameter estimators are to be obtained by the maximum penalized likelihood method of Firth [Firth, D., 1993, Bias reduction of maximum likelihood estimates. Biometrika, 80, 27–38]. Although this method achieves a reduction in bias, we illustrate that the remaining bias may be substantial for small experiments, and propose minimization of the integrated mean square error, based on Firths estimates, as a suitable criterion for design selection. This approach is used to find locally optimal designs for two support points.

Fractional Factorial Designs for the Detection of Interactions between Design and Noise Factors

In industrial experiments on both design (control) factors and noise factors aimed at improving the quality of manufactured products, designs are needed which afford independent estimation of all design×noise interactions in as few runs as possible, while allowing aliasing between those factorial effects of less interest. An algorithm for generating orthogonal fractional factorial designs of this type is described for factors at two levels. The generated designs are appropriate for experimenting on individual factors or for experimentation involving group screening of factors.

Assessing long-chain ω-3 polyunsaturated fatty acids: a tailored food-frequency questionnaire is better.

OBJECTIVE Polyunsaturated fatty acids (PUFAs), including the long-chain (LC) ω-3 PUFAs, are important for health. The aim was to assess if the Anti-Cancer Council of Victoria Dietary Questionnaire (ACC DQ) accurately determines PUFA intakes compared with the recently validated electronic PUFA food-frequency questionnaire (FFQ). METHODS Forty-one study volunteers were recruited from the local Illawarra region of New South Wales, Australia. The method of triads was used to determine validity coefficients by comparing the ACC DQ intakes against a 3-d weighed food record and appropriate blood biomarkers (erythrocytes and plasma fatty acids). These validity coefficients were subsequently compared with previously published validity coefficients from the PUFA FFQ. RESULTS Using erythrocytes as the biomarker, the electronic PUFA FFQ had much higher validity coefficients compared with the ACC DQ for eicosapentaenoic acid (0.92 versus 0.19), docosahexaenoic acid (0.69 versus 0.26), and total LC ω-3 PUFAs (0.78 versus 0.23), respectively, whereas ω-6 PUFAs were comparable. Using plasma as the biomarker, the electronic PUFA FFQ had much higher validity coefficients compared with the ACC DQ for α-linolenic acid (0.96 versus 0.49), eicosapentaenoic acid (0.87 versus 0.19), docosahexaenoic acid (0.64 versus 0.24), and total LC ω-3 PUFAs (0.73 versus 0.21), respectively, whereas ω-6 PUFAs were comparable. CONCLUSION The validated electronic PUFA FFQ is better suited to determine ω-3 PUFA intakes than the ACC DQ.

Crossover designs in the presence of carry-over effects from two factors

Experiments, used in the telecommunications industry and elsewhere, are considered that involve the simultaneous application of levels of two unrelated factors, treatments and stimuli, to each of several subjects in a succession of time periods. The existence is suspected of carry-over effects of treatments and stimuli, in the period immediately following the period of their application. Methods are given for the construction of separate sequences of treatments and of stimuli; these methods are based on the Latin squares of Williams and of Russell. In the resulting designs, the treatments and stimuli are either orthogonal or nearly orthogonal, and the coincidence of the direct and carry-over effects of each factor is either balanced or nearly balanced. The efficiencies of the designs are assessed by comparing the average variances of elementary contrasts in the levels of each factor with appropriate lower bounds.

The New Zealand PUFA Semiquantitative Food Frequency Questionnaire Is a Valid and Reliable Tool to Assess PUFA Intakes in Healthy New Zealand Adults

The health benefits of PUFA are well established. There is no valid tool or complete fatty acid database to assess PUFA intake in New Zealand (NZ). This study aimed to develop, validate, and test the reproducibility of a NZ-specific PUFA FFQ. A semiquantitative NZ PUFA FFQ was developed based on a validated Australian PUFA FFQ. The Australian fatty acid database was adapted to include NZ-specific data for major PUFA sources. Healthy participants from Auckland, NZ (n = 48) provided fasting blood samples for erythrocyte PUFA analysis, completed the NZ PUFA FFQ and a 3-d weighed food record (WFR), and repeated the NZ PUFA FFQ 3 mo later (n = 42). Relative validity was evaluated by assessing the triangular relationship among the NZ PUFA FFQ, WFR, and erythrocyte PUFA using the methods of triads [EPA, DHA, total omega-3 (n-3) long-chain (LC) PUFA only] and by comparing, correlating, cross-classifying into quintiles and assessing agreement using Bland-Altman plots of intakes between the NZ PUFA FFQ and WFR. Reproducibility was assessed by comparing and correlating intakes between repeat administrations of the NZ PUFA FFQ. The NZ PUFA FFQ effectively estimated EPA [ρ(QT) = 0.72 (95% CI: 0.49, 0.89)], DHA [ρ(QT) = 0.72 (95% CI: 0.53, 0.95)], and total (n-3) LCPUFA [ρ(QT) = 0.68 (95% CI: 0.47, 0.89)] intakes and was comparable with the WFR for other PUFA except docosapentaenoic acid. Repeated implementation of the NZ PUFA FFQ showed agreement for PUFA intakes. The NZ PUFA FFQ is a valid and reliable tool to measure PUFA intake in healthy NZ adults.

Theory & Methods: Resolvable Designs With Unequal Block Sizes

Resolvable block designs for v varieties in blocks of size k require v to be a multiple of k so that all blocks are of the same size. If a factorization of v is not possible then a resolvable design with blocks of unequal size is necessary. Patterson & Williams (1976) suggested the use of designs derived from α-designs and conjectured that such designs are likely to be very efficient in the class of resolvable designs with block sizes k and k– 1. This paper examines these derived designs and compares them with designs generated directly using an interchange algorithm. It concludes that the derived designs should be used when v is large, but that for small v they can be relatively inefficient.