Cryptic multiple hypotheses testing in linear models: overestimated effect sizes and the winner's curse

Fitting generalised linear models (GLMs) with more than one predictor has become the standard method of analysis in evolutionary and behavioural research. Often, GLMs are used for exploratory data analysis, where one starts with a complex full model including interaction terms and then simplifies by removing non-significant terms. While this approach can be useful, it is problematic if significant effects are interpreted as if they arose from a single a priori hypothesis test. This is because model selection involves cryptic multiple hypothesis testing, a fact that has only rarely been acknowledged or quantified. We show that the probability of finding at least one ‘significant’ effect is high, even if all null hypotheses are true (e.g. 40% when starting with four predictors and their two-way interactions). This probability is close to theoretical expectations when the sample size (N) is large relative to the number of predictors including interactions (k). In contrast, type I error rates strongly exceed even those expectations when model simplification is applied to models that are over-fitted before simplification (low N/k ratio). The increase in false-positive results arises primarily from an overestimation of effect sizes among significant predictors, leading to upward-biased effect sizes that often cannot be reproduced in follow-up studies (‘the winner's curse’). Despite having their own problems, full model tests and P value adjustments can be used as a guide to how frequently type I errors arise by sampling variation alone. We favour the presentation of full models, since they best reflect the range of predictors investigated and ensure a balanced representation also of non-significant results.     

A new approach to investigate the interactions between sediment transport and ecotoxicological processes during flood events

Purpose

The 'step-by step' principle was introduced into the European Union legislation on genetically modified organisms as a means to cope with uncertainty about environmental risks from the release of genetically modified organisms into the environment. The approval process is orientated along the stepwise reduction of containment which reflects a precautionary approach towards the risks of genetically modified organism release. Thus, the gradual reduction of containment should keep pace with the gradual generation of risk-related knowledge. This paper strives to clarify the meaning, legal status and practical importance of the principle. It also looks at whether non-European Union countries have adopted the principle as well, and how they practice it.

Methods

The article is based on research of the relevant legal texts, court cases and legal literature. In addition, a number of dossiers of applications for the European Union authorisation of release and placing on the market of genetically modified seed were analysed.

Results and conclusions

Although 'step-by-step' is not a precise legal rule it does have legal meaning as a principle guiding the risk assessment and management of genetically modified organism introduction into the environment. Assuming a process of gradual reduction of containment and scaling up of release ranging from closed systems via experimental release to cultivation the 'step-by-step principle' requires that the knowledge on environmental risks of genetically modified organisms should be generated on stages previous to the ones where the risk can result in damage. The analysis of the legislation of China, the United States of America and Brazil showed a differentiated approach towards the step-by-step principle.     

The role of the queen mandibular gland pheromone in honeybees (<Emphasis Type="Italic">Apis mellifera</Emphasis>): honest signal or suppressive agent?

Queen pheromones interfere with worker reproduction in social insects. However, there is still an unresolved question as to whether this pheromone acts as an “honest” signal for workers, giving a reliable indication of the queen’s reproductive value, or as a suppressive agent, inhibiting worker reproduction independent of the queen’s reproductive capacity. In honeybees (Apis mellifera), the queen’s mandibular gland secretion, a mix of fatty acids and some aromatic compounds, is crucial for regulating the reproductive division of labor in the colony inhibiting ovary development in workers. We quantified the mandibular gland secretions of virgin, drone-laying, and naturally mated queens using gas chromatography to test whether the queens’ mating, ovary activation, or the reproductive value for workers correlated with the composition of the secretion. Although the absolute amounts of the “queen substance” 9-oxo-2(E)-decenoic acid (9-ODA) were similar among the three groups, the proportions of 9-ODA decreased with increasing reproductive quality. Furthermore, the ratios of queen to worker compounds were similar in all three treatment groups, irrespective of the reproductive capacity. A multivariate analysis including all six compounds could not separate drone-laying queens from naturally mated ones, both with active ovaries but only the latter ensuring colony survival. We suggest that the mandibular gland pheromones are unlikely to function as reliable indicators of queen reproductive value and rather operate as an agent to suppress worker reproduction. This does not exclude the possibility that other “honest” pheromone signals exist in the honeybee colony, but these would have to arise from other semiochemicals, which could be produced by both the queen and the brood.     

Two years of oil disaster in the Arabian Gulf

In January–February 1991, about 1 million m³ of crude oil were released into the Arabian Gulf as a means of ecological warfare. A stretch of 644 km along the Saudi Arabian coast was heavily polluted over a width between a few meters and more than one kilometer. In four studies performed between March 1991 and April 1993 the state of representative areas of the affected coast was examined with the result that only minor changes were found in those oiled zones which are seldom inundated by the tides. Technical measures up to 1993 did not exceed the test level. A biological regeneration process is in progress which was not recognized as such in analyses of previous coastal oil pollution. Cyanobacterial mats, containing hydrocarbon-degrading bacteria, grow on oiled sediments which are regularly reached by tidal water. Shrinking and fracturing of the stable tar crusts starts a process of loosening and degrading of aged hydrocarbons.     

Relationship between the loading rate of inorganic mercury to aquatic ecosystems and dissolved gaseous mercury production and evasion

The purpose of our study was to test the hypothesis that dissolved gaseous mercury (DGM) production and evasion is directly proportional to the loading rate of inorganic mercury [Hg(II)] to aquatic ecosystems. We simulated different rates of atmospheric mercury deposition in 10-m diameter mesocosms in a boreal lake by adding multiple additions of Hg(II) enriched with a stable mercury isotope (²⁰²Hg). We measured DGM concentrations in surface waters and estimated evasion rates using the thin-film gas exchange model and mass transfer coefficients derived from sulfur hexafluoride (SF₆) additions. The additions of Hg(II) stimulated DGM production, indicating that newly added Hg(II) was highly reactive. Concentrations of DGM derived from the experimental Hg(II) additions (“spike DGM”) were directly proportional to the rate of Hg(II) loading to the mesocosms. Spike DGM concentrations averaged 0.15, 0.48 and 0.94 ng l⁻¹ in mesocosms loaded at 7.1, 14.2, and 35.5 μg Hg m⁻² yr⁻¹, respectively. The evasion rates of spike DGM from these mesocosms averaged 4.2, 17.2, and 22.3 ng m⁻² h⁻¹, respectively. The percentage of Hg(II) added to the mesocosms that was lost to the atmosphere was substantial (33–59% over 8 weeks) and was unrelated to the rate of Hg(II) loading. We conclude that changes in atmospheric mercury deposition to aquatic ecosystems will not change the relative proportion of mercury recycled to the atmosphere.     

可持续建筑部门的理念与手段

本文概览了伍珀塔尔研究所为建筑部门确立可持续性目标,以及为实现这些目标开创途径所采取的多种方法.资源生产力是覆盖建筑物整个生命周期的考虑(MIPS).为了迈向更大的可持续性,COMPASS理念把单一公司或工业部门的环境、经济和社会各方面因素综合考虑.这些方法中的每一种都有所作为,再根据不同类型的研究,为公司和政策制定者提出建议.多边利益相关者程序可以用来推动建筑部门的全面可持续发展,最终把相关的理念明确地推进到微型和中等级别上.     

Attenuation reactions in a multiple contaminated aquifer in Bitterfeld (Germany)

Large-scale contaminated sites with multiple contaminants in the groundwater present a challenge to risk assessment and remediation. Attenuation reactions take place in the subsurface and act to contain contaminants, but must be thoroughly investigated on a site-specific basis. Field data from monitoring wells at a contaminated industrial site in Bitterfeld, Germany, are presented and analyzed for evidence of the prevalent biodegradation reactions. The groundwater in the Tertiary aquifer is contaminated with large quantities of chlorinated aliphatic compounds, in addition to chlorobenzenes and BTEX. In this strictly anaerobic environment, geochemical indications for several microbial processes were found, including methanogenesis, sulfate and iron reduction as well as reductive dechlorination of the chlorinated hydrocarbons. Direct evidence for the latter degradation reaction was observed along the flowpath due to the appearance of intermediates and an increase in the degree of dechlorination.     

Formation of chloromethoxybenzaldehyde during composting of organic household waste

Standardized household waste was mixed with different litter amendments, straw, leaves, hardwood shavings, softwood shavings, paper, and sphagnum peat, resulting in six compost mixtures. In addition non-amended household waste was composted. Composting was done in small rotatable bins and compost samples were taken on a regular basis until day 590. Extraction and analysis of wet compost samples showed no evidence for the presence of chloroorganic compounds. Drying and re-wetting of compost samples, however, revealed that chloromethoxybenzaldehyde (CMBA) was formed in all composts at concentrations varying between 5.6 and 73.4 microg kg(-1) dry matter. CMBA was not present in the original materials. During composting, there was a clear positive relation between formation of CMBA and microbial activity, as indicated by C losses and temperature. Formation took place during the most intensive phase of composting when C losses were highest. Under anaerobic conditions, however, which prevailed initially in the non-amended compost, no CMBA was formed. Calculation of total amounts of CMBA in composts revealed that there was a small decrease during storage in the hardwood, peat, and softwood composts. However, all composts contained CMBA after 590 days. The mean concentration was 33.4 microg kg(-1) dry matter (s.d. = 21.9). Possible biocidal effects of composts when used in cultivation may be explainable by the presence of natural toxic compounds formed during composting.