Jørund Gåsemyr

On an adaptive version of the Metropolis-Hastings algorithm with independent proposal distribution

In this paper, we present a general formulation of an algorithm, the adaptive independent chain (AIC), that was introduced in a special context in Gasemyr et al. [Methodol. Comput. Appl. Probab. 3 (2001)]. The algorithm aims at producing samples from a specific target distribution Π, and is an adaptive, non-Markovian version of the Metropolis-Hastings independent chain. A certain parametric class of possible proposal distributions is fixed, and the parameters of the proposal distribution are updated periodically on the basis of the recent history of the chain, thereby obtaining proposals that get ever closer to Π. We show that under certain conditions, the algorithm produces an exact sample from Π in a finite number of iterations, and hence that it converges to II. We also present another adaptive algorithm, the componentwise adaptive independent chain (CAIC), which may be an alternative in particular in high dimensions. The CAIC may be regarded as an adaptive approximation to the Gibbs sampler updating parametric approximations to the conditionals of II.

Simulation based analysis and an application to an offshore oil and gas production system of the Natvig measures of component importance in repairable systems

In the present paper the Natvig measures of component importance for repairable systems, and its extended version are analyzed for two three-component systems and a bridge system. The measures are also applied to an offshore oil and gas production system. According to the extended version of the Natvig measure a component is important if both by failing it strongly reduces the expected system uptime and by being repaired it strongly reduces the expected system downtime. The results include a study of how different distributions affect the ranking of the components. All numerical results are computed using discrete event simulation. In a companion paper [Huseby AB, Eide KA, Isaksen SL, Natvig B, Gasemyr, J. Advanced discrete event simulation methods with application to importance measure estimation. 2009, submitted for publication] the advanced simulation methods needed in these calculations are described.

Does atenolol differ from other β-adrenergic blockers?

BackgroundA recent meta-analysis of drug effects in patients with hypertension claims that all β-adrenergic blockers are equally effective but less so than other antihypertensive drugs. Published comparisons of the β-adrenergic blocker atenolol and non-atenolol β-adrenergic blockers indicate different effects on death rates, arrhythmias, peripheral vascular resistance and prognosis post myocardial infarction, all in disfavour of atenolol. In keeping with these findings, the data presented in the meta-analysis indicate that atenolol is less effective than the non-atenolol β-adrenergic blockers both when compared with placebo and with other antihypertensive drugs. These findings were not, however, statistically significant.MethodsWe performed an additional analysis with a Bayesian statistical method in order to make further use of the published data.ResultsOur calculations on the clinical data in the meta-analysis showed 13% lower risk (risk ratio 0.87) of myocardial infarction among hypertensive patients taking non-atenolol β-adrenergic blockers than among hypertensive patients taking atenolol. The 90 % credibility interval ranged from 0.75 to 0.99, thereby indicating statistical significance. The probability of at least 10% lower risk (risk ratio ≤ 0.90), which could be considered to be of clinical interest, was 0.69.ConclusionTaken together with the other observations of differences in effects, we conclude that the claim that all β-adrenergic blockers are inferior drugs for hypertensive patients should be rejected. Atenolol is not representative of the β-adrenergic blocker class of drugs as a whole and is thus not a suitable drug for comparisons with other antihypertensive drugs in terms of effect. The non-atenolol β-adrenergic blockers should thus continue to be fundamental in antihypertensive drug treatments.

The posterior distribution of the parameters of component lifetimes based on autopsy data in a shock model

In this paper we consider a binary, monotone system whose component states are dependent through the possible occurrence of independent common shocks, i.e. shocks that destroy several components at once. The individual failure of a component is also thought of as a shock. Such systems can be used to model common cause failures in reliability analysis. The system may be a technological one, or a human being. It is observed until it fails or dies. At this instant, the set of failed components and the failure time of the system are noted. The failure times of the components are not known. These are the so-called autopsy data of the system. For the case of independent components, i.e. no common shocks, Meilijson (1981), Nowik (1990), Antoine et al. (1993) and GTsemyr (1998) discuss the corresponding identifiability problem, i.e. whether the component life distributions can be determined from the distribution of the observed data. Assuming a model where autopsy data is known to be enough for identifia bility, Meilijson (1994) goes beyond the identifiability question and into maximum likelihood estimation of the parameters of the component lifetime distributions based on empirical autopsy data from a sample of several systems. He also considers life-monitoring of some components and conditional life-monitoring of some other. Here a corresponding Bayesian approach is presented for the shock model. Due to prior information one advantage of this approach is that the identifiability problem represents no obstacle. The motivation for introducing the shock model is that the autopsy model is of special importance when components can not be tested separately because it is difficult to reproduce the conditions prevailing in the functioning system. In Gasemyr & Natvig (1997) we treat the Bayesian approach to life-monitoring and conditional life- monitoring of components

Comparing Effects of Biologic Agents in Treating Patients with Rheumatoid Arthritis: A Multiple Treatment Comparison Regression Analysis.

Rheumatoid arthritis patients have been treated with disease modifying anti-rheumatic drugs (DMARDs) and the newer biologic drugs. We sought to compare and rank the biologics with respect to efficacy. We performed a literature search identifying 54 publications encompassing 9 biologics. We conducted a multiple treatment comparison regression analysis letting the number experiencing a 50% improvement on the ACR score be dependent upon dose level and disease duration for assessing the comparable relative effect between biologics and placebo or DMARD. The analysis embraced all treatment and comparator arms over all publications. Hence, all measured effects of any biologic agent contributed to the comparison of all biologic agents relative to each other either given alone or combined with DMARD. We found the drug effect to be dependent on dose level, but not on disease duration, and the impact of a high versus low dose level was the same for all drugs (higher doses indicated a higher frequency of ACR50 scores). The ranking of the drugs when given without DMARD was certolizumab (ranked highest), etanercept, tocilizumab/ abatacept and adalimumab. The ranking of the drugs when given with DMARD was certolizumab (ranked highest), tocilizumab, anakinra, rituximab, golimumab/ infliximab/ abatacept, adalimumab/ etanercept. Still, all drugs were effective. All biologic agents were effective compared to placebo, with certolizumab the most effective and adalimumab (without DMARD treatment) and adalimumab/ etanercept (combined with DMARD treatment) the least effective. The drugs were in general more effective, except for etanercept, when given together with DMARDs.

Clinical efficacies of antihypertensive drugs.

Objective—According to published data, the ability to prevent various hypertension-related events differs between the various antihypertensive drug groups. Although absolute drug effects differ among studies, relative drug effects could be considered constant. We therefore explored the possibility of drawing statistically valid conclusions about the differences in clinical efficacy between various drug groups by doing an overview of published data. Design—We made a meta-analysis with a Bayesian fixed effect model in which we related the drug effects to the effects of placebo drugs. We selected 27 clinical trials from the literature according to specific criteria, including results from studies reporting the effects of the newer drugs when tested against diuretics and, β-blockers, and from studies in which diuretics and, β-blockers had been tested against placebo. We calculated the posterior probability distributions of the relative effects of angiotensin-converting enzyme (ACE) inhibitors vs calcium antagonists with three different endpoints: stroke, coronary disease and heart failure with point estimates of effects and with 95% credibility intervals. As an intermediate step in this procedure we obtained similar information about the effects of the three groups of active drugs, ACE inhibitors, calcium antagonists and diuretics or, β-blockers, tested against placebo. For coronary disease we also tested calcium antagonists against diuretics or, β-blockers. Results—ACE inhibitors and calcium antagonists have an almost identical ability to prevent stroke in hypertensive individuals with a risk ratio (RR) of 1.04. On the other hand, calcium antagonists reduce coronary disease by only 8% relative to placebo. When ACE inhibitors and calcium antagonists are compared with the Bayesian method, the outcome is a 14% difference in favor of the ACE inhibitors to prevent coronary disease, with a credibility interval almost reaching identity. Nor do calcium antagonists do as well as diuretics or, β-blockers in this respect, RR = 1.12 with 95% credibility interval 1.01-1.24. All the tested drug groups have a profound preventive effect on the occurrence of heart failure when given to hypertensive patients, showing reductions of 42-54%. When ACE inhibitors are compared with calcium antagonists RR = 0.79, with a credibility interval 0.65-0.95. Conclusion—There is statistically an indisputable difference between ACE inhibitors and calcium antagonists in respect of effects on coronary disease and heart failure when treating hypertensive individuals, ACE inhibitors being more efficacious. There are no differences in the effect on stroke. Moreover, ,β-blockers or diuretics are also superior to calcium antagonists in preventing coronary events.

Meta-Regression Analysis of Paroxetine Clinical Trial Data: Does Reporting Scale Matter?

Objectives: It is unknown to which degree the effect of antidepressant drugs are related to baseline degree of depression, dose level, patients age, or type of questionnaire used. We explored this for paroxetine. Methods: We used placebo-controlled published and unpublished randomized, double-blind, clinical trials of paroxetine that included moderate to severely depressed patients in an outpatient setting. We specified random-effect models for the Hamilton 17-item and Hamilton 21-item studies separately and jointly. Results: Among 35 studies retrieved, we considered 26 appropriate for a pooled analysis. Paroxetine (placebo) was given to 2958 (2123) patients. We found that the effects of paroxetine, the differences between score reduction in drug versus placebo group, are smaller in Hamilton 17 (3.8%) than in Hamilton 21 studies (7.0%). The mean difference is 3.2% (95% confidence interval, 0.94%-5.42%), statistically significant by meta-regression analysis. Treatment effects did not change with mean age of patients, early or late studies, baseline score value, or maximal daily dose. Conclusions: We forward 2 hypotheses for explanation. The Hamilton 21 studies had better selection of patients, thereby smaller effect of regression to the mean than the Hamilton 17 studies, meaning the Hamilton 21 studies reveal true somewhat higher treatment effects. Alternatively, the study groups contained some patients with psychotic symptoms tested for with the Hamilton 21-item questionnaire and thereby becoming decisive for the outcome. If so, paroxetine would have an antipsychotic effect. This is in accordance with some experimental and clinical observations.

Advanced Discrete Event Simulation Methods with Application to Importance Measure Estimation in Reliability

In the present paper we use discrete event simulation in order to analyze a binary monotone system of repairable components. Asymptotic statistical properties of such a system, e.g., the asymptotic system availability and component criticality, can easily be estimated by running a single discrete event simulation on the system over a sufficiently long time horizon, or by working directly on the stationary component availabilities. Sometimes, however, one needs to estimate how the statistical properties of the system evolve over time. In such cases it is necessary to run many simulations to obtain a stable curve estimate. At the same time one needs to store much more information from each simulation. A crude approach to this problem is to sample the system state at fixed points of time, and then use the mean values of the states at these points as estimates of the curve. Using a sufficiently high sampling rate a satisfactory estimate of the curve can be obtained. Still, all information about the process between the sampling points is thrown away. To handle this issue, we propose an alternative sampling procedure where we utilize process data between the sampling points as well. This simulation method is particularly useful when estimating various kinds of component importance measures for repairable systems. As explained in [11] such measures can often be expressed as weighted integrals of the time-dependent Birnbaum measure of importance. By using the proposed simulation methods, stable estimates of the Birnbaum measure as a function of time are obtained. Combined with the appropriate weight function the importance measures of interest can be estimated.

A Comparison of two Sequential Metropolis-Hastings Algorithms with Standard Simulation Techniques in Bayesian Inference in Reliability Models Involving the Generalized Gamma Distribution

In this paper we consider the generalized gamma distribution as introduced in Gåsemyr and Natvig (1998). This distribution enters naturally in Bayesian inference in exponential survival models with left censoring. In the paper mentioned above it is shown that the weighted sum of products of generalized gamma distributions is a conjugate prior for the parameters of component lifetimes, having autopsy data in a Marshall-Olkin shock model. A corresponding result is shown in Gåsemyr and Natvig (1999) for independent, exponentially distributed component lifetimes in a model with partial monitoring of components with applications to preventive system maintenance. A discussion in the present paper strongly indicates that expressing the posterior distribution in terms of the generalized gamma distribution is computationally efficient compared to using the ordinary gamma distribution in such models. Furthermore, we present two types of sequential Metropolis-Hastings algorithms that may be used in Bayesian inference in situations where exact methods are intractable. Finally these types of algorithms are compared with standard simulation techniques and analytical results in arriving at the posterior distribution of the parameters of component lifetimes in special cases of the mentioned models. It seems that one of these types of algorithms may be very favorable when prior assessments are updated by several data sets and when there are significant discrepancies between the prior assessments and the data.

A multiple treatment comparison meta-analysis of monoamine oxidase type B inhibitors for Parkinson's disease

To the best of our knowledge, there are no systematic reviews or meta‐analyses that compare rasagiline, selegiline and safinamide. Therefore, we aimed to perform a drug class review comparing all available monoamine oxidase type B (MAO‐B) inhibitors in a multiple treatment comparison.