Tests multiples simulés et tests de normalité basés sur plusieurs moments dans les modèles de régression

Cet article illustre l’applicabilité des méthodes de rééchantillonnage dans le cadre des tests multiples (simultanés), pour divers problèmes économétriques. Les hypothèses simultanées sont une conséquence habituelle de la théorie économique, de sorte que le contrôle de la probabilité de rejet de combinaisons de tests est un problème que l’on rencontre fréquemment dans divers contextes économétriques et statistiques. À ce sujet, on sait que le fait d’ignorer le caractère conjoint des hypothèses multiples peut faire en sorte que le niveau de la procédure globale dépasse considérablement le niveau désiré. Alors que la plupart des méthodes d’inférence multiple sont conservatrices en présence de statistiques non-indépendantes, les tests que nous proposons visent à contrôler exactement le niveau de signification. Pour ce faire, nous considérons des critères de test combinés proposés initialement pour des statistiques indépendantes. En appliquant la méthode des tests de Monte Carlo, nous montrons comment ces méthodes de combinaison de tests peuvent s’appliquer à de tels cas, sans recours à des approximations asymptotiques. Après avoir passé en revue les résultats antérieurs sur ce sujet, nous montrons comment une telle méthodologie peut être utilisée pour construire des tests de normalité basés sur plusieurs moments pour les erreurs de modèles de régression linéaires. Pour ce problème, nous proposons une généralisation valide à distance finie du test asymptotique proposé par Kiefer et Salmon (1983) ainsi que des tests combinés suivant les méthodes de Tippett et de Pearson-Fisher. Nous observons empiriquement que les procédures de test corrigées par la méthode des tests de Monte Carlo ne souffrent pas du problème de biais (ou sous-rejet) souvent rapporté dans cette littérature – notamment contre les lois platikurtiques – et permettent des gains sensibles de puissance par rapport aux méthodes combinées usuelles.

Probe-level linear model fitting and mixture modeling results in high accuracy detection of differential gene expression

Affiliation: Institut de recherche en immunologie et en cancérologie, Université de Montréal

Comparaison de différents traitements psychologiques dans l'endométriose

Cette thèse avait pour but premier d’évaluer la douleur chronique endométriosique et ses concomitants (dépression, anxiété et stress), les conséquences de la douleur sur le physique, les activités et le travail, sur la relation maritale et les séquelles sur la qualité de vie chez des participantes souffrant de douleurs pelviennes chroniques diagnostiquées endométriose (laparoscopie). En deuxième lieu, il s’agissait d’évaluer et de comparer l’efficacité des techniques psychologiques de contrôle de la douleur (Hypnose, Cognitif-behavioral) en ajout aux traitements médicaux à un groupe contrôle (Attention thérapeute). L’échantillon était composé de 60 femmes réparties aléatoirement soit à l’un des deux groupes de traitement ou au groupe contrôle. Les instruments de mesure étaient tous des questionnaires déjà traduits en français et validés pour la population francophone québécoise. Des tests du khi-carré ont été effectués pour les variables nominales et des analyses de variances (ANOVA) ont été faites pour les variables continues. Dans des modèles ANOVA estimant l’effet du traitement, du temps et de leur interaction, une différence significative (effet de Groupe ou traitement) a été trouvée pour les variables suivantes : Douleur (McGill :composante évaluative p = 0.02), au moment « présent » de l’Échelle visuelle analogique (EVA, p = 0.05) et dans l’Échelle de Qualité de vie (douleur, p = 0,03) ainsi qu’à la dimension Fonctionnement social de cette dernière échelle (SF-36; p = 0,04). En comparant les données en pré et post-traitement, des résultats significatifs au niveau du Temps ont aussi été mis en évidence pour les variables suivantes : Douleur McGill: Score total, (p = 0,03), Affective (p = 0,04), Évaluative (p = 0,01); Douleur (ÉVA) moment Fort (p < 0,0005), Dépression (p = 0,005), Anxiété (situationnelle/état (p = 0,002), Anxiété/trait (p < 0,001), Stress (p = 0, 003) ainsi que pour quatre composantes de la Qualité de vie (Fonctionnement social, (p = 0,05), Vitalité (p = 0,002), Douleur, (p = 0,003) et Changement de la santé (p < 0,001) et ceci pour les trois groupes à l’exception du groupe Hypnose sur cette dernière variable. Des effets d’Interaction (Groupe X Temps) sont ressortis sur les variables « Conséquences physiques » de la douleur mais sur la dimension « Activités » seulement (p = 0,02), sur l’anxiété situationnelle (État : p = 0,007). Un effet d’interaction se rapprochant de la signification (p = 0,08) a aussi été analysé pour la variable Fonctionnement social (SF-36). L’étude montre une légère supériorité quant au traitement Cognitif-behavioral pour l’anxiété situationnelle, pour le Fonctionnement social et pour la douleur mesurée par le SF-36. L’étude présente des forces (groupe homogène, essai clinique prospectif, répartition aléatoire des participantes et groupe contrôle) mais aussi des lacunes (faible échantillon et biais potentiels reliés à l’expérimentateur et à l’effet placebo). Toute future étude devrait tenir compte de biais potentiels quant au nombre d’expérimentateur et inclure un groupe placebo spécifique aux études à caractère psychologique. Une future étude devrait évaluer le schème cognitif « catastrophisation » impliqué dans la douleur, les traits de personnalité des participantes ainsi que le rôle du conjoint. De plus, des techniques psychologiques (entrevues motivationnelles) récentes utilisées dans plusieurs études devraient aussi être prises en considérations. Tout de même des résultats significatifs offrent des pistes intéressantes pour un essai clinique comportant un échantillon plus élevé et pour un suivi à long terme.

Influence of Weapon Types on the Patterns and Outcomes of Violent Encounters

Rapport de stage présenté à l'école de Criminologie de la Faculté des arts et sciences en vue de l'obtention du grade de Maître ès sciences (M.Sc.) en criminologie option analyse criminologique

Essais en économetrie et économie de l'éducation

Cette thèse est organisée en trois chapitres. Les deux premiers s'intéressent à l'évaluation, par des méthodes d'estimations, de l'effet causal ou de l'effet d'un traitement, dans un environnement riche en données. Le dernier chapitre se rapporte à l'économie de l'éducation. Plus précisément dans ce chapitre j'évalue l'effet de la spécialisation au secondaire sur le choix de filière à l'université et la performance. Dans le premier chapitre, j'étudie l'estimation efficace d'un paramètre de dimension finie dans un modèle linéaire où le nombre d'instruments peut être très grand ou infini. L'utilisation d'un grand nombre de conditions de moments améliore l'efficacité asymptotique des estimateurs par variables instrumentales, mais accroit le biais. Je propose une version régularisée de l'estimateur LIML basée sur trois méthodes de régularisations différentes, Tikhonov, Landweber Fridman, et composantes principales, qui réduisent le biais. Le deuxième chapitre étend les travaux précédents, en permettant la présence d'un grand nombre d'instruments faibles. Le problème des instruments faibles est la consequence d'un très faible paramètre de concentration. Afin d'augmenter la taille du paramètre de concentration, je propose d'augmenter le nombre d'instruments. Je montre par la suite que les estimateurs 2SLS et LIML régularisés sont convergents et asymptotiquement normaux. Le troisième chapitre de cette thèse analyse l'effet de la spécialisation au secondaire sur le choix de filière à l'université. En utilisant des données américaines, j'évalue la relation entre la performance à l'université et les différents types de cours suivis pendant les études secondaires. Les résultats suggèrent que les étudiants choisissent les filières dans lesquelles ils ont acquis plus de compétences au secondaire. Cependant, on a une relation en U entre la diversification et la performance à l'université, suggérant une tension entre la spécialisation et la diversification. Le compromis sous-jacent est évalué par l'estimation d'un modèle structurel de l'acquisition du capital humain au secondaire et de choix de filière. Des analyses contrefactuelles impliquent qu'un cours de plus en matière quantitative augmente les inscriptions dans les filières scientifiques et technologiques de 4 points de pourcentage.

The impact of early intra-articular corticosteroid injections on the outcome of oligoarticular juvenile idiopathic arthritis

Contexte Un objectif important de la prise en charge de l'arthrite juvénile oligoarticulaire serait d'altérer le cours de la maladie à l'aide d'une thérapie hâtive. Nous avons étudié l'effet des injections intra-articulaires de corticostéroïdes hâtives sur les chances d'atteindre un décompte d'articulation active de zéro et une maladie inactive. Méthode Les données démographiques, cliniques et thérapeutiques des patients avec oligoarthrite juvénile enrôlés dans une étude prospective longitudinale pancanadienne ont été collectées pendant 2 ans. Une injection hâtive était définie comme étant reçue dans les 3 premiers mois suivant le diagnostic. Les équations d'estimation généralisées ont été utilisées pour l'analyse statistique. Résultats Trois cent dix patients ont été inclus. Cent onze (35.8%) ont reçu une injection hâtive. Ces derniers avaient une maladie plus active lors de l'entrée dans l'étude. Les patients exposés à une injection hâtive avaient une chance similaire d'obtenir un décompte d'articulation active de zéro, OR 1.52 (IC95% 0.68-3.37), p=0.306 mais étaient significativement moins à risque d'avoir une maladie inactive, OR 0.35 (IC95% 0.14-0.88), p=0.026. Interprétation Dans cette cohorte de 310 patients avec oligoarthrite juvénile, les injections hâtives de corticostéroïdes n'ont pas mené à une probabilité plus élevée d'atteindre un décompte d'articulation active de zéro ou une maladie inactive. Des problématiques méthodologiques intrinsèques à l'utilisation de données observationnelles pour fins d'estimation d'effets thérapeutiques auraient pu biaiser les résultats. Nous ne pouvons affirmer avec certitude que les injections hâtives n'améliorent pas le décours de la maladie. Des études prospectives adressant les limitations soulevées seront requises pour clarifier la question.

Studies on Advanced Oxidation Processes for the Removal of Acetamiprid from Wastewater

Pollution of water with pesticides has become a threat to the man, material and environment. The pesticides released to the environment reach the water bodies through run off. Industrial wastewater from pesticide manufacturing industries contains pesticides at higher concentration and hence a major source of water pollution. Pesticides create a lot of health and environmental hazards which include diseases like cancer, liver and kidney disorders, reproductive disorders, fatal death, birth defects etc. Conventional wastewater treatment plants based on biological treatment are not efficient to remove these compounds to the desired level. Most of the pesticides are phyto-toxic i.e., they kill the microorganism responsible for the degradation and are recalcitrant in nature. Advanced oxidation process (AOP) is a class of oxidation techniques where hydroxyl radicals are employed for oxidation of pollutants. AOPs have the ability to totally mineralise the organic pollutants to CO2 and water. Different methods are employed for the generation of hydroxyl radicals in AOP systems. Acetamiprid is a neonicotinoid insecticide widely used to control sucking type insects on crops such as leafy vegetables, citrus fruits, pome fruits, grapes, cotton, ornamental flowers. It is now recommended as a substitute for organophosphorous pesticides. Since its use is increasing, its presence is increasingly found in the environment. It has high water solubility and is not easily biodegradable. It has the potential to pollute surface and ground waters. Here, the use of AOPs for the removal of acetamiprid from wastewater has been investigated. Five methods were selected for the study based on literature survey and preliminary experiments conducted. Fenton process, UV treatment, UV/ H2O2 process, photo-Fenton and photocatalysis using TiO2 were selected for study. Undoped TiO2 and TiO2 doped with Cu and Fe were prepared by sol-gel method. Characterisation of the prepared catalysts was done by X-ray diffraction, scanning electron microscope, differential thermal analysis and thermogravimetric analysis. Influence of major operating parameters on the removal of acetamiprid has been investigated. All the experiments were designed using central compoiste design (CCD) of response surface methodology (RSM). Model equations were developed for Fenton, UV/ H2O2, photo-Fenton and photocatalysis for predicting acetamiprid removal and total organic carbon (TOC) removal for different operating conditions. Quality of the models were analysed by statistical methods. Experimental validations were also done to confirm the quality of the models. Optimum conditions obtained by experiment were verified with that obtained using response optimiser. Fenton Process is the simplest and oldest AOP where hydrogen peroxide and iron are employed for the generation of hydroxyl radicals. Influence of H2O2 and Fe2+ on the acetamiprid removal and TOC removal by Fenton process were investigated and it was found that removal increases with increase in H2O2 and Fe2+ concentration. At an initial concentration of 50 mg/L acetamiprid, 200 mg/L H2O2 and 20 mg/L Fe2+ at pH 3 was found to be optimum for acetamiprid removal. For UV treatment effect of pH was studied and it was found that pH has not much effect on the removal rate. Addition of H2O2 to UV process increased the removal rate because of the hydroxyl radical formation due to photolyis of H2O2. An H2O2 concentration of 110 mg/L at pH 6 was found to be optimum for acetamiprid removal. With photo-Fenton drastic reduction in the treatment time was observed with 10 times reduction in the amount of reagents required. H2O2 concentration of 20 mg/L and Fe2+ concentration of 2 mg/L was found to be optimum at pH 3. With TiO2 photocatalysis improvement in the removal rate was noticed compared to UV treatment. Effect of Cu and Fe doping on the photocatalytic activity under UV light was studied and it was observed that Cu doping enhanced the removal rate slightly while Fe doping has decreased the removal rate. Maximum acetamiprid removal was observed for an optimum catalyst loading of 1000 mg/L and Cu concentration of 1 wt%. It was noticed that mineralisation efficiency of the processes is low compared to acetamiprid removal efficiency. This may be due to the presence of stable intermediate compounds formed during degradation Kinetic studies were conducted for all the treatment processes and it was found that all processes follow pseudo-first order kinetics. Kinetic constants were found out from the experimental data for all the processes and half lives were calculated. The rate of reaction was in the order, photo- Fenton>UV/ H2O2>Fenton> TiO2 photocatalysis>UV. Operating cost was calculated for the processes and it was found that photo-Fenton removes the acetamiprid at lowest operating cost in lesser time. A kinetic model was developed for photo-Fenton process using the elementary reaction data and mass balance equations for the species involved in the process. Variation of acetamiprid concentration with time for different H2O2 and Fe2+ concentration at pH 3 can be found out using this model. The model was validated by comparing the simulated concentration profiles with that obtained from experiments. This study established the viability of the selected AOPs for the removal of acetamiprid from wastewater. Of the studied AOPs photo- Fenton gives the highest removal efficiency with lowest operating cost within shortest time.

Effects of fertilizer type and rate on partitioning of soil organic matter into pools of different stability

Type and rate of fertilizers influence the level of soil organic carbon (Corg) and total nitrogen (Nt) markedly, but the effect on C and N partitioning into different pools is open to question. The objectives of the present work were to: (i) quantify the impact of fertilizer type and rate on labile, intermediate and passive C and N pools by using a combination of biological, chemical and mathematical methods; (ii) explain previously reported differences in the soil organic matter (SOM) levels between soils receiving farmyard manure with or without biodynamic preparations by using Corg time series and information on SOM partitioning; and (iii) quantify the long-term and short-term dynamics of SOM in density fractions and microbial biomass as affected by fertilizer type and rate and determine the incorporation of crop residues into labile SOM fractions. Samples were taken from a sandy Cambisol from the long-term fertilization trial in Darmstadt, Germany, founded in 1980. The nine treatments (four field replicates) were: straw incorporation plus application of mineral fertilizer (MSI) and application of rotted farmyard manure with (DYN) or without (FYM) addition of biodynamic preparations, each at high (140 – 150 kg N ha-1 year-1; MSIH, DYNH, FYMH), medium (100 kg N ha-1 year-1; MSIM, DYNM, FYMM) and low (50 – 60 kg N ha-1 year-1; MSIL, DYNL, FYML) rates. The main findings were: (i) The stocks of Corg (t ha-1) were affected by fertilizer type and rate and increased in the order MSIL (23.6), MSIM (23.7), MSIH (24.2) < FYML (25.3) < FYMM (28.1), FYMH (28.1). Stocks of Nt were affected in the same way (C/N ratio: 11). Storage of C and N in the modelled labile pools (turnover times: 462 and 153 days for C and N, respectively) were not influenced by the type of fertilizer (FYM and MSI) but depended significantly (p ≤ 0.05) on the application rate and ranged from 1.8 to 3.2 t C ha 1 (7 – 13% of Corg) and from 90 to 140 kg N ha-1 (4-5% of Nt). In the calculated intermediate pool (C/N ratio 7), stocks of C were markedly higher in FYM treatments (15-18 t ha-1) compared to MSI treatments (12-14 t ha-1). This showed that differences in SOM stocks in the sandy Cambisol induced by fertilizer rate may be short-lived in case of changing management, but differences induced by fertilizer type may persist for decades. (ii) Crop yields, estimated C inputs (1.5 t ha-1 year-1) with crop residue, microbial bio¬mass C (Cmic, 118 – 150 mg kg-1), microbial biomass N (17 – 20 mg kg-1) and labile C and N pools did not differ significantly between FYM and DYN treatments. However, labile C increased linearly with application rate (R2 = 0.53) from 7 to 11% of Corg. This also applied for labile N (3.5 to 4.9% of Nt). The higher contents of Corg in DYN treatments existed since 1982, when the first sampling was conducted for all individual treatments. Contents of Corg between DYN and FYM treatments con-verged slightly since then. Furthermore, at least 30% of the difference in Corg was located in the passive pool where a treatment effect could be excluded. Therefore, the reported differences in Corg contents existed most likely since the beginning of the experiment and, as a single factor of biodynamic agriculture, application of bio-dynamic preparations had no effect on SOM stocks. (iii) Stocks of SOM, light fraction organic C (LFOC, ρ ≤ 2.0 g cm-3), light fraction organic N and Cmic decreased in the order FYMH > FYML > MSIH, MSIL for all sampling dates in 2008 (March, May, September, December). However, statistical significance of treatment effects differed between the dates, probably due to dif-ferences in the spatial variation throughout the year. The high proportion of LFOC on total Corg stocks (45 – 55%) highlighted the importance of selective preservation of OM as a stabilization mechanism in this sandy Cambisol. The apparent turnover time of LFOC was between 21 and 32 years, which agreed very well with studies with substantially longer vegetation change compared to our study. Overall, both approaches; (I) the combination of incubation, chemical fractionation and simple modelling and (II) the density fractionation; provided complementary information on the partitioning of SOM into pools of different stability. The density fractionation showed that differences in Corg stocks between FYM and MSI treatments were mainly located in the light fraction, i.e. induced by higher recalcitrance of the organic input in the FYM treatments. Moreover, the use of the combination of biological, chemical and mathematical methods indicated that effects of fertilizer rate on total Corg and Nt stocks may be short-lived, but that the effect of fertilizer type may persist for longer time spans in the sandy Cambisol.

Förderung des Leseverständnisses durch Lesestrategien: Eine Interventionsstudie in der Grundschule

Anliegen dieser Dissertation ist die Untersuchung des Einflusses eines Lesestrategietrainings auf die Entwicklung des Leseverständnisses in den Klassen 2 und 4. In einer experimentellen Studie mit Prä-Post-Test-Design wurde die Wirksamkeit eines 25 Sitzungen umfassenden Trainings der Strategien Vorhersagen, kleinschrittiges Wiederholen und Zusammenfassen den Effekten einer lesefernen Kontrollbedingung gegenübergestellt. Die Lesestrategien wurden peergestützt vermittelt, d. h. in Dyaden aus einem leseschwächeren und einem lesestärkeren Kind. In drei Teilstudien wurden Fragestellungen zur differenzierten Analyse der Trainingswirksamkeit untersucht: (1) Wird die Trainingswirksamkeit durch die Effizienz der Worterkennungsprozesse beeinflusst?, (2) Kann die Entwicklung der Leseflüssigkeit durch das Lesestrategietraining gefördert werden? und (3) Können leseschwache Kinder von der Zusammenarbeit mit lesestärkeren Tutor(inn)en hinsichtlich der Verbesserung ihres Leseverständnisses profitieren?. Die Ergebnisse dieser Dissertation sprechen dafür, dass das eingesetzte peergestützte Lesestrategietraining das Leseverständnis und die Leseflüssigkeit von Zweit- und Viertklässler(inne)n unter bestimmten Voraussetzungen positiv beeinflussen konnte. Die Leseleistungen vor dem Training, die Effizienz der Worterkennungsprozesse und die Rolle im dyadischen Lernen erwiesen sich als relevante Einflussfaktoren für die Wirksamkeit des Strategietrainings. Zweitklässler(innen), die aufgrund guter Prä-Test Leseleistungen die Tutor(inn)enrolle erhielten, konnten ihr Leseverständnis gegenüber Kindern mit gleichen Leseleistungen in der Kontrollbedingung signifikant steigern. Leseschwache Zweitklässler(innen) hingegen schienen nur bei (relativ) effizienten Worterkennungsprozessen die Lesestrategien zur Steigerung ihres globalen Leseverständnisses nutzen zu können, wobei sie keinen Zugewinn aus der dyadischen Zusammenarbeit ziehen konnten. Bei ineffizienten Worterkennungsprozessen hatte das Strategietraining negative Auswirkungen auf das allgemeine Leseverständnis. Anders in Klasse 4: Kinder, die aufgrund unterdurchschnittlicher Leseleistungen im Prä-Test als Tutand(inn)en am Training teilnahmen, verbesserten ihr Leseverständnis und konnten dabei von der Zusammenarbeit mit lesestarken Tutor(inn)en profitieren. Für die Tutor(inn)en in Klasse 4 zeigte sich kein Effekt des Strategietrainings gegenüber der Kontrollgruppe. In beiden Klassenstufen stellte sich im Verlauf des Strategietrainings eine schnellere Entwicklung der Leseflüssigkeit der schwachen Leser(innen) ein. Zusammengefasst erzielte das Training die größten Effekte für gute Leser(innen) in Klasse 2 und schwache Leser(innen) in Klasse 4 – beide Gruppen konnten ihr Leseverständnis praktisch bedeutsam gegenüber Kindern mit gleichen Leseleistungen in der Kontrollbedingung steigern.

Measuring and testing for the systemically important financial institutions

This paper analyzes the measure of systemic importance ∆CoV aR proposed by Adrian and Brunnermeier (2009, 2010) within the context of a similar class of risk measures used in the risk management literature. In addition, we develop a series of testing procedures, based on ∆CoV aR, to identify and rank the systemically important institutions. We stress the importance of statistical testing in interpreting the measure of systemic importance. An empirical application illustrates the testing procedures, using equity data for three European banks.

Empowering IDP with SMS: a randomized controlled trial in Bogotá

We carried out a randomized controlled trial in Bogotá, the recipient of Colombia´s highest number of internally displaced people (IDP), to assess whether the use of SMS to communicate eligibility to social benefits fosters the welfare of victimized internal refugees. Only a fraction of IDP are elegible to benefits. We inform eligibility via SMS to a random half of IDP-households who are, and estimate the Local Average Treatment Effect of the text message on the knowledge of the benefits available tothe displaced population. We show that while on the average treated households know their rights better than controls, a more disaggregate analysis suggest that there is variation of awareness across benefits. The intervention was overall successful in empowering IDP and the use of SMS should be widened as a social policy instrument. However our results suggest that text messages should be complemented with other communication strategies, yet to be evaluated.

Numerical Response of Breeding Birds Following Experimental Selection Harvesting in Northern Hardwood Forests

Silvicultural treatments have been shown to alter the composition of species assemblages in numerous taxa. However, the intensity and persistence of these effects have rarely been documented. We used a before-after, control-impact (BACI) paired design, i.e., five pairs of 25-ha study plots, 1-control and 1-treated plot, to quantify changes in the density of eight forest bird species in response to selection harvesting over six breeding seasons, one year pre- and five years postharvest. Focal species included mature forest associates, i.e., Northern Parula (Setophaga americana) and Black-throated Green Warbler (Setophaga virens), forest generalists, i.e., Yellow-bellied Sapsucker (Sphyrapicus varius) and Swainson’s Thrush (Catharus ustulatus), early-seral specialists, i.e., Mourning Warbler (Geothlypis philadelphia) and Chestnut-sided Warbler (Setophaga pensylvanica), species associated with shrubby forest gaps, i.e., Black-throated Blue Warbler (Setophaga caerulescens), and mid-seral species, i.e., American Redstart (Setophaga ruticilla). As predicted, we found a negative numerical response to the treatment in the Black-throated Green Warbler, no treatment effect in the Yellow-bellied Sapsucker, and a positive treatment effect in early-seral specialists. We only detected a year effect in the Northern Parula and the American Redstart. There was evidence for a positive treatment effect on the Swainson’s Thrush when the regeneration started to reach the pole stage, i.e., fifth year postharvest. These findings suggest that selection harvesting has the potential to maintain diverse avian assemblages while allowing sustainable management of timber supply, but future studies should determine whether mature-forest associates can sustain second- and third-entry selection harvest treatments.

Meta-analysis of individual patient data from randomised trials: a review of methods used in practice

Background: Meta-analyses based on individual patient data (IPD) are regarded as the gold standard for systematic reviews. However, the methods used for analysing and presenting results from IPD meta-analyses have received little discussion. Methods We review 44 IPD meta-analyses published during the years 1999–2001. We summarize whether they obtained all the data they sought, what types of approaches were used in the analysis, including assumptions of common or random effects, and how they examined the effects of covariates. Results: Twenty-four out of 44 analyses focused on time-to-event outcomes, and most analyses (28) estimated treatment effects within each trial and then combined the results assuming a common treatment effect across trials. Three analyses failed to stratify by trial, analysing the data is if they came from a single mega-trial. Only nine analyses used random effects methods. Covariate-treatment interactions were generally investigated by subgrouping patients. Seven of the meta-analyses included data from less than 80% of the randomized patients sought, but did not address the resulting potential biases. Conclusions: Although IPD meta-analyses have many advantages in assessing the effects of health care, there are several aspects that could be further developed to make fuller use of the potential of these time-consuming projects. In particular, IPD could be used to more fully investigate the influence of covariates on heterogeneity of treatment effects, both within and between trials. The impact of heterogeneity, or use of random effects, are seldom discussed. There is thus considerable scope for enhancing the methods of analysis and presentation of IPD meta-analysis.

Statistical design and analysis of pharmacogenetic trials

Pharmacogenetic trials investigate the effect of genotype on treatment response. When there are two or more treatment groups and two or more genetic groups, investigation of gene-treatment interactions is of key interest. However, calculation of the power to detect such interactions is complicated because this depends not only on the treatment effect size within each genetic group, but also on the number of genetic groups, the size of each genetic group, and the type of genetic effect that is both present and tested for. The scale chosen to measure the magnitude of an interaction can also be problematic, especially for the binary case. Elston et al. proposed a test for detecting the presence of gene-treatment interactions for binary responses, and gave appropriate power calculations. This paper shows how the same approach can also be used for normally distributed responses. We also propose a method for analysing and performing sample size calculations based on a generalized linear model (GLM) approach. The power of the Elston et al. and GLM approaches are compared for the binary and normal case using several illustrative examples. While more sensitive to errors in model specification than the Elston et al. approach, the GLM approach is much more flexible and in many cases more powerful. Copyright © 2005 John Wiley & Sons, Ltd.

A practical comparison of group-sequential and adaptive designs

Sequential methods provide a formal framework by which clinical trial data can be monitored as they accumulate. The results from interim analyses can be used either to modify the design of the remainder of the trial or to stop the trial as soon as sufficient evidence of either the presence or absence of a treatment effect is available. The circumstances under which the trial will be stopped with a claim of superiority for the experimental treatment, must, however, be determined in advance so as to control the overall type I error rate. One approach to calculating the stopping rule is the group-sequential method. A relatively recent alternative to group-sequential approaches is the adaptive design method. This latter approach provides considerable flexibility in changes to the design of a clinical trial at an interim point. However, a criticism is that the method by which evidence from different parts of the trial is combined means that a final comparison of treatments is not based on a sufficient statistic for the treatment difference, suggesting that the method may lack power. The aim of this paper is to compare two adaptive design approaches with the group-sequential approach. We first compare the form of the stopping boundaries obtained using the different methods. We then focus on a comparison of the power of the different trials when they are designed so as to be as similar as possible. We conclude that all methods acceptably control type I error rate and power when the sample size is modified based on a variance estimate, provided no interim analysis is so small that the asymptotic properties of the test statistic no longer hold. In the latter case, the group-sequential approach is to be preferred. Provided that asymptotic assumptions hold, the adaptive design approaches control the type I error rate even if the sample size is adjusted on the basis of an estimate of the treatment effect, showing that the adaptive designs allow more modifications than the group-sequential method.