Misleading Associations Research Articles

Elementary classical mechanics and the principle of the Composition of Causes

In this paper, I explore whether elementary classical mechanics adheres to the Principle of Composition of Causes as Mill claimed and as certain contemporary authors still seem to believe. Among other things, I provide a proof that if one reads Mill’s description of the principle literally (as I think many do), it does not hold in any general sense. In addition, I explore a separate notion of Composition of Causes and note that it too does not hold in elementary classical mechanics. Among the major morals is that there is no utility to describing classical mechanics in terms of Composition of Causes. This is both because the stated principles do not hold and because when one describes what actually does hold in classical mechanics in terms of the Composition of Causes, one introduces misleading associations that can generate errors just as claimed by Russell (Mysticism and logic, 1981).

Principles of statistical mechanics: The energy duality

We argue that statistical mechanics of systems with relaxation implies breaking the energy function of systems into two having different transformation rules. With this duality the energy approach incorporates the generalized vortex forces. We show general theorems and their implications and apply the approach to the particle confinement in fields of rotational symmetry. Misconceptions of extensive use of the quasienergy and generalized thermodynamic potential theories are exposed.

CARDIOVASCULAR AND SURVIVAL PARADOXES IN DIALYSIS PATIENTS: Misleading Associations between Cholesterol and Vascular Outcomes in Dialysis Patients: The Need for Randomized Trials

AbstractHigher cholesterol is strongly associated with an increased risk of coronary heart disease (CHD) in nonrenal populations, so the lack of a clear positive association between total cholesterol and mortality among dialysis patients is unexpected. This review of prospective studies of the association between total cholesterol and mortality among dialysis patients suggests that there is a negative association at below average cholesterol levels and a flat or weakly positive association at higher levels. In nonrenal populations total cholesterol is not positively associated with vascular causes of death other than CHD, so the lack of a strongly positive association at above average cholesterol concentrations in dialysis patients may be explained by the high proportion of deaths due to non‐CHD vascular causes. The observation of a negative association between total cholesterol and mortality at below average cholesterol concentrations in some studies is consistent with confounding by both vascular and nonvascular morbidity (i.e., reverse causality). We argue that evaluating the importance of cholesterol for vascular disease risk in dialysis patients can only be achieved through the eradication of confounding by randomization, and that ongoing trials of cholesterol‐lowering therapy will provide a definitive answer to this question.

A Model-Based Frequency Constraint for Mining Associations from Transaction Data

Mining frequent itemsets is a popular method for finding associated items in databases. For this method, support, the co-occurrence frequency of the items which form an association, is used as the primary indicator of the associations's significance. A single, user-specified support threshold is used to decided if associations should be further investigated. Support has some known problems with rare items, favors shorter itemsets and sometimes produces misleading associations. In this paper we develop a novel model-based frequency constraint as an alternative to a single, user-specified minimum support. The constraint utilizes knowledge of the process generating transaction data by applying a simple stochastic mixture model (the NB model) which allows for transaction data's typically highly skewed item frequency distribution. A user-specified precision threshold is used together with the model to find local frequency thresholds for groups of itemsets. Based on the constraint we develop the notion of NB-frequent itemsets and adapt a mining algorithm to find all NB-frequent itemsets in a database. In experiments with publicly available transaction databases we show that the new constraint provides improvements over a single minimum support threshold and that the precision threshold is more robust and easier to set and interpret by the user.

An integrated model of the behavioural dimensions of industrial buyer‐seller relationships

PurposeResearch on the behavioural aspects of buyer‐seller relationships, although sizeable, is too heterogeneous and fragmented to yield complete and conclusive insights as to the inter‐relationships of the basic parameters involved. This article attempts to put together extant knowledge on the subject under an integrated conceptual model comprising ten key behavioural constructs.Design/methodology/approachA total of 24 theoretically‐anchored hypotheses are developed, indicating possible positive or negative associations among the constructs of the model. Based on input received from 122 producers of industrial goods, the proposed model is empirically tested using structural equation modelling.FindingsOf the hypothesized associations examined, 16 were found to be statistically significant and in the right direction, two were significant but in the opposite direction, while the remaining six were not validated. The results confirm most of the findings of previous research on the subject, while some fresh insights on the interrelationships of the constructs used are also revealed.Research limitations/implicationsThe results of the study have serious implications for industrial marketers, organizational buyers, management consultants, and business educators, who may use the empirically tested model as a diagnostic and monitoring tool in guiding business relationships in the proper direction.Originality/valueThe article concludes that an integrative approach to studying the behavioural aspects of industrial buyer‐seller relationships can provide a more realistic understanding of the constructs at work, compared with a partial one which may conceal some of the effects of one construct over others and/or show misleading associations among constructs.

The fundamental flaw in obesity research

The basic problem with comparative diet trials is our inability to measure what people eat. All conventional instruments depend on subjects' reports. Most trials lack independent biochemical, physiological or genetic measures of intake. So, we do not know if subjects actually follow the diets being tested and compared. We can assess weight gain/loss, but we fail in a fundamental scientific requirement, accurately measuring the independent variable in a causal experiment. Worse, we know most subjects under-report their energy intake and its components, the obese especially. The problem is compounded by attempts to show diets' effects on other risk factors, like triglycerides. Researchers seek to correlate two variables, without having accurately measured one of them, producing misleading associations. The consequence is we do not know if the results of any current diet trials are valid or reliable. Developing rigorous measures of food intake is the highest priority in obesity research. That involves improvements in technology as well as science. We need: (1) biomarkers of intake for energy, macro- and micro-nutrients and other food components relevant to weight gain/loss; (2) field measuring instruments that are cheap, rapid, painless, non-intrusive and self-administerable; and (3) electronic data transmission systems that preclude subjects' ability to misreport.

Comments on the Papers by Guthrie and Sheppard, Best et al., Chambers and Steel and Darby et al.

Ben Armstrong (London School of Hygiene and Tr-opical Medicine) An ecological study is defined in epidemiology as 'A study in which the units of analysis are populations or groups of people rather than individuals' (Last, 1988). Ecological studies have long been introduced to students of epidemiology as occupying bottom place in the hierarchy of designs, barely above case reports. Most damning of all, students learn that whereas other designs might suffer from 'biases' and 'confounding', which can be controlled, ecological studies are said to suffer from the 'ecological fallacy', which appears to have no remedy. This is usually the only context in which students of epidemiology encounter this especially disapproving term. The ecological fallacy is usually demonstrated with a reference to Durkheim's example (cited by Greenland and Robins (1994)), of association between suicide rates and proportions of Protestants in Swiss Cantons-possibly due to Catholics committing suicide in Cantons dominated by Protestants. This with some examples of some obviously misleading associations between average levels of risk factors and outcomes across countries generally suffices to fix firmly in students' minds a deep-rooted lifelong suspicion of ecological studies. Published methodological work beginning as a trickle in the early 1980s and reaching a steady stream by the early 1990s has sought to clarify more carefully conditions under which ecological studies of disease yield useful results, and to improve methods of design and analysis to make this more likely. Emphasis has generally remained on the limitations, but some strengths have also been noted. A greater robustness to measurement error bias is an example. Several epidemiological developments have encouraged methodological work in this area. First, it has been observed that some putative disease risk factors, notably diet, vary much more between communities than between individuals within the same community (Rose, 1985). Could the advantages of this greater contrast outweigh the disadvantages of the ecological design? Second, hybrid or 'semiecological' designs have become more common. In these some explanatory variables can be linked to the outcome at an individual level, and others are available only at the group level. An example is the 'six-cities' study of air pollution and health, in which air pollution was measured at the city level, but smoking, other personal risk factors and health outcomes were measured in individuals (Dockery et al., 1993). Finally, small scale geographical referencing has enabled ecological studies in which the units of analysis are very small areas (Elliott et al., 1992). To what extent do semniecological and small area studies escape the limitations that are expected of ecological studies? Useful recent reviews from an epidemiological perspective of the methodological work before this issue include Walter (1991), Richardson (1992), Greenland (1992), Greenland and Robins (1994) and Morgenstern (1998). Although ecological studies often suffer from problems that are not intrinsic to the ecological design (aggregated data), such as omitted or poorly measured confounders, most of the literature is concerned with biases directly due to aggregation, usually called ecological bias. Ecological bias is a particular problem in epidemiology because under the non-linearity of most epidemiological models conditions for the absence of ecological bias are much rarer than under the linear models that are popular in other contexts (Greenland and Robins, 1994). This set of papers extends the understanding of ecological analyses in several complementary ways. Guthrie and Sheppard explore the potential of design and analysis techniques introduced as 'aggregate data' methods by Prentice and Sheppard to overcome limitations of ecological studies. In this approach models at an individual level are used to induce models at the group level, which can then be fitted to group level data to estimate individual level parameters. For non-linear models the group level models are typically functions of variances and covariances of explanatory variables, rather than the means used in conventional ecological analysis. As these are rarely available from routine sources, additional data collection is required, at least on a subsample of subjects. By comparing aggregate data estimators with conventional ecological methods in simulations, Guthrie and Sheppard confirm that these methods

Use of Parents, Sibs, and Unrelated Controls for Detection of Associations between Genetic Markers and Disease

Detecting the association between genetic markers and complex diseases can be a critical first step toward identification of the genetic basis of disease. Misleading associations can be avoided by choosing as controls the parents of diseased cases, but the availability of parents often limits this design to early-onset disease. Alternatively, sib controls offer a valid design. A general multivariate score statistic is presented, to detect the association between a multiallelic genetic marker locus and affection status; this general approach is applicable to designs that use parents as controls, sibs as controls, or even unrelated controls whose genotypes do not fit Hardy-Weinberg proportions or that pool any combination of these different designs. The benefit of this multivariate score statistic is that it will tend to be the most powerful method when multiple marker alleles are associated with affection status. To plan these types of studies, we present methods to compute sample size and power, allowing for varying sibship sizes, ascertainment criteria, and genetic models of risk. The results indicate that sib controls have less power than parental controls and that the power of sib controls can be increased by increasing either the number of affected sibs per sibship or the number of unaffected control sibs. The sample-size results indicate that the use of sib controls to test for associations, by use of either a single-marker locus or a genomewide screen, will be feasible for markers that have a dominant effect and for common alleles having a recessive effect. The results presented will be useful for investigators planning studies using sibs as controls.

Relation Between Inbred and Hybrid Traits in Maize1

We conducted correlation studies between plant and ear traits of inbred lines and single‐cross hybrids of maize (Zea mays L.) developed from Iowa Stiff Stalk Synthetic. We included 160 S7 lines and 320 single‐cross hybrids produced among the lines by use of the cross‐classification mating design. The 160 S7 lines were partitioned into 20 sets of eight lines to produce the 320 single‐cross hybrids (16 hybrids per set). Objectives of our study were to determine the relation between plant and ear traits of the inbred lines and their single‐cross hybrids and to determine the relation of plant and ear traits of the inbred lines with yield of the single‐cross hybrids. Two procedures were used: (1) correlations of inbred lines themselves with means of hybrids that had one inbred line in common and (2) correlations of the means of two inbred lines with their specific hybrids. Simple and multiple correlations were computed for each set of eight inbred lines and 16 single‐cross hybrids and combined across the 20 sets.Simple correlations of plant and ear traits of inbred lines and single‐cross hybrids were small in all instances for both procedures when combined across sets. The coefficient of determination (r2) was less than 0.16 in all instances. Correlations for individual sets for all traits ranged from significantly (P ≤ 0.01) negative to significantly positive. Sampling error was high among individual sets, indicating that small sample sizes can give misleading associations of traits. Multiple correlations of plant and ear traits of inbred lines with yield of singlecross hybrids were only 0.23 (Procedure 1) and 0.21 (Procedure 2). Our results showed that plant and ear traits of inbred lines were not good indicators of their performance in single‐cross hybrids. Correlations of yield of inbred lines with yields of single‐cross hybrids were 0.09 and 0.11. It seems that yield testing of inbred lines in single‐cross hybrids is the only procedure for determining the potential usefulness of an inbred line in single‐cross hybrids. Visual selection of morphological traits in inbred lines will not insure development of higher yielding single‐cross yields.

Causes

In an essay on ‘The Notion of Cause’ reprinted in Mysticism and Logic (and which constituted the presidential address to the Aristotelian Society in the year 1912) Russell argued ‘that the word ‘cause” is so inextricably bound up with misleading associations as to make its complete extrusion from the philosophical vocabulary desirable’.1 His argument here to the effect that ‘cause’ is not a central concept in science, as philosophers have thought it, is reminiscent of Norman Campbell's statement in Physics: The Elements (1920) and in What is Science? (1921).