Statistical Inconsistency Research Articles

How useful and reliable are construction statistics?

The Pearce Report of 2003 drew attention to potential shortcomings in some of the existing statistics that were used to analyse the UK construction sector. The current study examines the broader statistical issues in greater depth to identify shortcomings that could cause some of the findings presented by Pearce, and similar studies, to need qualification and possible revision. Problems with reliable and accurate data collection and statistical analysis include defining the scope and coverage of the industry; measuring industry outputs and their allocation across different types of activity; identifying construction firms; and measuring capital formation and capital stock, inconsistencies in employment statistics and labour market variables, discrepancies in measuring productivity, and the lack of international comparison. Needed improvements include a wider definition of construction output and employment activities, a continuous review of the usefulness and reliability of the published construction statistics, and the resolution of the discrepancies between different registers. Industry stakeholders also need to become more engaged in the process of statistical revision and use. The creation of an international forum could improve the quality of statistics and deal with the more technical aspects of international comparisons. Le Rapport Pearce de 2003 a attiré l'attention sur les insuffisances potentielles de certaines statistiques existantes utilisées pour analyser le secteur de la construction au Royaume-Uni. L'étude actuelle examine la question des statistiques au sens large en l'approfondissant afin de recenser ces lacunes pour que les résultats présentés par Pearce (et dans d'autres études) soient qualifiés et, éventuellement révisés. Les problèmes liés à la collecte de données fiables et précises et à l'analyse statistique incluent la définition de la portée de ces études ainsi que la couverture de l'industrie, la mesure des données fournies par l'industrie et leur affectation aux différents types d'activités; l'identification des entreprises de construction et la mesure de la formation du capital et du capital social, les incohérences constatées dans les statistiques de l'emploi et dans les variables du marché du travail, les écarts dans la mesure de la productivité et le manque de comparaison à l'échelon international. Il convient, notamment, de définir plus largement les résultats fournis par le secteur de la construction et par les activités d'emploi, de revoir en continu l'utilité et la fiabilité des statistiques publiées par le secteur de la construction et de résoudre les écarts entre les différents registres. Les parties prenantes de cette industrie doivent également s'engager davantage dans le processus de révision et d'utilisation des statistiques. La création d'un forum international pourrait améliorer la qualité des statistiques et traiter les aspects plus techniques des comparaisons à l'échelon international. Mots clés: performances de la construction, collecte de données, emploi, résultats, productivité, fiabilité, mesure des ressources, statistiques

Finding the truth behind photographs

In the analog world, photographic images have generally been accepted as ‘proof of occurrence’ of the depicted events. But today, the creation and manipulation of images is made simple by digital processing tools that are widely available and easily obtained. As a result, we can no longer take the authenticity of any images, analog or digital, for granted. This is especially true when they are used as photographic evidence by law enforcement agencies. In this context, the field of image forensics is concerned with uncovering underlying facts about an image: by using its techniques, we attempt to provide authoritative answers to several questions about the content and source of an image. Is this an ‘original’, or was it created by cut and paste operations from different images? Was it captured by a camera manufactured by vendor X or vendor Y? Did it originate from camera X as claimed? At time Y? At location Z? Does this image truly represent the original scene or was it digitally altered to deceive the viewer: is this coffee stain a re-colored blood stain? These are just a few of the questions faced routinely by investigators and others in law enforcement. However, existing techniques are often insufficient to provide authoritative answers. Although digital watermarks have been proposed as a tool to verify the authenticity of images, the overwhelming majority of images captured today do not have one. This situation is likely to continue for the foreseeable future, so in the absence of their widespread adoption, we believe it is imperative to develop techniques that can generate definitive statements about the origin, veracity, and nature of digital images. The past few years have seen a growth of research in this area. Work in the field has focused mainly on solving two types of problems: image source verification and tamper detection. The aim of the former is to determine through what means a given image was generated (e.g., digital-camera, computer graphics, scanner, etc.) and associate it with a class of sources that have common characteristics, or match it to a specific source. Meanwhile, the goal of tamper detection is to determine whether a given image has undergone any form of modification or processing since it was initially captured. Our research at the Information System and Internet Security (ISIS) laboratory at Polytechnic University, NY, has resulted in the development of many techniques that address these two types of problems. We briefly summarize this work here. When the source of an image is identified as being a digital camera, the ability to identify its make and model requires an understanding of the image formation process. Although many of the details of the camera pipeline are considered proprietary information by manufacturers, the basic structure remains the same in all digital cameras. In our initial approach to the problem, we identified a number of relevant features. These included energy in sub-spectral bands; higher-order statistics of sub-band coefficients, which can be used to demonstrate statistical inconsistencies between natural and unnatural images; image quality metrics; and inter-band correlations. Also, deviations from the ‘gray world’ assumption— which states that, in a natural image with a sufficient amount of color variation, the values of each of the basic color components of the image (red, green, and blue) should average to gray—were identified as possible indicators of alteration. Based on these features, we used a classification-based technique to capture the characteristics of each camera examined. This approach was also applied to cell phone camera images with reasonable success. We further enhanced our results by detecting interpolation artifacts in smooth and non-smooth parts of the image. These were introduced using a color-filter array, which necessitates a de-mosaicing operation. Our approach to the detection of image-tampering stems from the idea that a digitally altered image would have undergone one or more image-processing operations (e.g., scaling, rotation, blurring/enhancement, brightness/contrast adjustments). For this purpose, we designed classifiers that can distinguish between images that both have and have not been processed us-

Phylogenomics

▪ Abstract The continuous flow of genomic data is creating unprecedented opportunities for the reconstruction of molecular phylogenies. Access to whole-genome data means that phylogenetic analysis can now be performed at different genomic levels, such as primary sequences and gene order, allowing for reciprocal corroboration of the results. We critically review the different kinds of phylogenomic methods currently available, paying particular attention to method reliability. Our emphasis is on methods for the analysis of primary sequences because these are the most advanced. We discuss the important issue of statistical inconsistency and show how failing to fully capture the process of sequence evolution in the underlying models leads to tree reconstruction artifacts. We suggest strategies for detecting and potentially overcoming these problems. These strategies involve the development of better models, the use of an improved taxon sampling, and the exclusion of phylogenetically misleading data.

The Constraining Power of International Treaties: Theory and Methods

We acknowledge the contribution of von Stein (2005) in calling attention to the very real problem of selection bias in estimating treaty effects. Nonetheless, we dispute both von Stein's theoretical and empirical conclusions. Theoretically, we contend that treaties can both screen and constrain simultaneously, meaning that findings of screening do nothing to undermine the claim that treaties constrain state behavior as well. Empirically, we question von Stein's estimator on several grounds, including its strong distributional assumptions and its statistical inconsistency. We then illustrate that selection bias does not account for much of the difference between Simmons's (2000) and von Stein's (2005) estimated treaty effects, and instead reframe the problem as one of model dependency. Using a preprocessing matching step to reduce that dependency, we then illustrate treaty effects that are both substantively and statistically significant—and that are quite close in magnitude to those reported by Simmons.

Parsimony overcomes statistical inconsistency with the addition of more data from the same gene.

Many authors have demonstrated that the parsimony method of phylogenetic analysis can fail to estimate phylogeny accurately under certain conditions when data follow a model that stipulates homogeneity of the evolutionary process. These demonstrations further show that no matter how much data are added, parsimony will forever exhibit this statistical inconsistency if the additional data have the same distributional properties as the original data. This final component-that the additional data must follow the same distribution as the original data-is crucial to the demonstration. Recent simulations show, however, that if data evolve heterogeneously, parsimony can perform consistently. Here we show, using natural data, that parsimony can overcome inconsistency if new data from the same gene are added to an analysis already exhibiting a condition indistinguishable from inconsistency.

Delayed versus immediate exercises following surgery for breast cancer: a systematic review

Seroma formation, wound healing and fluid drainage are a concern for both surgeons and patients. Excessive fluid production can result in seroma formation, and inadequate drainage of seromas is known to cause infection, pain, discomfort and longer periods of hospitalisation. Postoperative exercises given to maintain movement of the arm are believed to increase the amount of fluid production following surgery. This review aimed to determine whether a program of delayed exercises reduces the risk of seroma formation, fluid loss and hospital stay, without loss of arm movement. A systematic review. RCTs of early versus delayed shoulder mobilisation after surgery in females with breast cancer were included in the review. Outcomes. One or more measurements of shoulder range of motion, wound complications, fluid drainage volumes and incidence of seroma formation. Design. Randomised controlled trials, control group of delayed exercise/mobilisation. Validity assessment was carried out using a data extraction form based on the CONSORT statement. Study characteristics recorded include sample size, intervention, control, period of exercise delay, surgical procedure and conclusions drawn. Data synthesis was carried out using random effects and weighted mean differences to test for heterogeneity and combined effects. 12 RCTs were included in the review of which 6 were included for meta-analysis. Delaying exercises significantly decreases seroma formation (OR=0.4; 95%CI 0.2-0.5; p=0.00001). No significant differences were found for drainage volume or hospital stay. Current evidence from RCTs supports the use of a delayed program of arm exercises to reduce seroma formation. Clinical and statistical inconsistencies between studies did not allow any conclusions to be drawn regarding the effects of delayed exercises on fluid drainage, hospital stay and immediate or long term ability to move the arm.

Should we be worried about long-branch attraction in real data sets? Investigations using metazoan 18S rDNA

Although long-branch attraction (LBA) is frequently cited as the cause of anomalous phylogenetic groupings, few examples of LBA involving real sequence data are known. We have found several cases of probable LBA by analyzing subsamples from an alignment of 18S rDNA sequences for 133 metazoans. In one example, maximum parsimony analysis of sequences from two rotifers, a ctenophore, and a polychaete annelid resulted in strong support for a tree grouping two “long-branch taxa” (a rotifer and the ctenophore). Maximum-likelihood analysis of the same sequences yielded strong support for a more biologically reasonable “rotifer monophyly” tree. Attempts to break up long branches for problematic subsamples through increased taxon sampling reduced, but did not eliminate, LBA problems. Exhaustive analyses of all quartets for a subset of 50 sequences were performed in order to compare the performance of maximum likelihood, equal-weights parsimony, and two additional variants of parsimony; these methods do differ substantially in their rates of failure to recover trees consistent with well established, but highly unresolved phylogenies. Power analyses using simulations suggest that some incorrect inferences by maximum parsimony are due to statistical inconsistency and that when estimates of central branch lengths for certain quartets are very low, maximum-likelihood analyses have difficulty recovering accepted phylogenies even with large amounts of data. These examples demonstrate that LBA problems can occur in real data sets, and they provide an opportunity to investigate causes of incorrect inferences.

Do mistakes matter?

Now hear this. Non-steroidal anti-inflammatory drugs are better than opioids for relieving renal colic (p 1401). Now hear another thing. Enteral nutrition produces a quicker recovery from acute pancreatitis than parentral nutrition (p 1407). And another two. Epidural analgesia does not increase risk of caesarean section (p 1410), nor does H pylori eradication have any effect on heartburn or reflux (p 1417). But can you believe what you hear? A study recently published on Biomedcentral (http://www.biomedcentral.com/) found statistical inconsistencies in 38% of papers in Nature and 25% in the BMJ . Emili Garcia-Berthou and Carles Alcaraz examined …

Distances that perfectly mislead.

Given a collection of discrete characters (e.g., aligned DNA sites, gene adjacencies), a common measure of distance between taxa is the proportion of characters for which taxa have different character states. Tree reconstruction based on these (uncorrected) distances can be statistically inconsistent and can lead to trees different from those obtained using character-based methods such as maximum likelihood or maximum parsimony. However, in these cases the distance data often reveal their unreliability by some deviation from additivity, as indicated by conflicting support for more than one tree. We describe two results that show how uncorrected (and miscorrected) distance data can be simultaneously perfectly additive and misleading. First, multistate character data can be perfectly compatible and define one tree, and yet the uncorrected distances derived from these characters are perfectly treelike (and obey a molecular clock), only for a completely different tree. Second, under a Markov model of character evolution a similar phenomenon can occur; not only is there statistical inconsistency using uncorrected distances, but there is no evidence of this inconsistency because the distances look perfectly treelike (this does not occur in the classic two-parameter Felsenstein zone). We characterize precisely when uncorrected distances are additive on the true (and on a false) tree for four taxa. We also extend this result to a more general setting that applies to distances corrected according to an incorrect model.

Statistical procedures for calculating the residual virulence of Brucella abortus strain 19 (S19) and Brucella melitensis strain Rev 1 vaccines in mice: theoretical basis and practical applications.

Regular control of the biological quality of live Brucella abortus strain 19 (S19) and B. melitensis strain Rev 1 vaccines is essential for the successful management of ruminant brucellosis in affected countries. The reference procedures recommended by the OIE (World organisation for animal health) and the European Pharmacopoeia include the determination of residual virulence, expressed as the recovery time 50 (RT50), of the tested (problem) vaccine in a reference mouse model compared with the RT50 of the corresponding reference strains in the same assay. The underlying statistical procedure applied is based on a parallel line assay and a classical probit model. In practice, the currently recommended procedure for calculating the RT50 is based on a graphical method which has never been described in detail. This paper provides a full description of this graphical method with the aim of making the technique comprehensible and accessible to all interested biologists. The procedure is somewhat cumbersome and very few laboratories apply the OIE and European Pharmacopoeia recommendations on a regular basis. Moreover, since this reference graphical method shows some statistical inconsistencies, a dedicated internet interface has been developed to perform RT50 calculations and is now available free of charge on the web (www.afssa.fr/interne/rev2.html).

Revisiting the Factor Structure of the Agoraphobic Cognitions Questionnaire and Body Sensations Questionnaire: A Confirmatory Factor Analysis Study

The twin scales, Agoraphobic Cognitions Questionnaire and Body Sensations Questionnaire (D. L. Chambless, G. C. Caputo, P. Bright, & R. Gallagher, 1984), have been used frequently in the area of anxiety, and in particular, panic disorder with agoraphobia. The present study employed confirmatory factor analysis to investigate the factor structure of the scales in a sample of Australian patients with anxiety disorder. Model testing indicated that neither the original nor the modified factor structures offered a good fit to the data. In addition, confirmatory factor analyses highlighted statistical inconsistencies in the scales. Except for a few, most of the items were problematic. This indicated a need for possible revision of the scales. Moreover, until further research is conducted, researchers and clinicians are recommended to use the scales with caution.

A Post-Mortem Analysis of Opinion and Exit Polls in West Bengal in May 2001

Statewide assembly election was held in May 2001 in West Bengal, where an alliance of leftist political parties has been in power since 1977. All opinion and exit polls conducted in the state predicted a closely contested election with a very slim projected majority for the ruling alliance, if at all. The actual outcome was another landslide victory for them. This paper reviews these opinion and exit poll results and shows certain statistical inconsistencies in their projected figures. A brief analysis of 2001 election results reveals some statistically significant evidence for the presence of anti-incumbancy factor acting against the ruling alliance candidates. This analysis also indicates that there was no significant decrease in the fragmentation of opposition votes in 2001 election compared to the most recent assembly election in 1996 and the most recent parliamentary election in 1999.

Error, bias, and long-branch attraction in data for two chloroplast photosystem genes in seed plants.

Sequences of two chloroplast photosystem genes, psaA and psbB, together comprising about 3,500 bp, were obtained for all five major groups of extant seed plants and several outgroups among other vascular plants. Strongly supported, but significantly conflicting, phylogenetic signals were obtained in parsimony analyses from partitions of the data into first and second codon positions versus third positions. In the former, both genes agreed on a monophyletic gymnosperms, with Gnetales closely related to certain conifers. In the latter, Gnetales are inferred to be the sister group of all other seed plants, with gymnosperms paraphyletic. None of the data supported the modern "anthophyte hypothesis," which places Gnetales as the sister group of flowering plants. A series of simulation studies were undertaken to examine the error rate for parsimony inference. Three kinds of errors were examined: random error, systematic bias (both properties of finite data sets), and statistical inconsistency owing to long-branch attraction (an asymptotic property). Parsimony reconstructions were extremely biased for third-position data for psbB. Regardless of the true underlying tree, a tree in which Gnetales are sister to all other seed plants was likely to be reconstructed for these data. None of the combinations of genes or partitions permits the anthophyte tree to be reconstructed with high probability. Simulations of progressively larger data sets indicate the existence of long-branch attraction (statistical inconsistency) for third-position psbB data if either the anthophyte tree or the gymnosperm tree is correct. This is also true for the anthophyte tree using either psaA third positions or psbB first and second positions. A factor contributing to bias and inconsistency is extremely short branches at the base of the seed plant radiation, coupled with extremely high rates in Gnetales and nonseed plant outgroups.

Commentary

Thus confounding is equated with noncollapsibility (or change-in-parameter estimate). The second, comparabil/ty-based definition views the effect of exposure as a comparison of the group's outcome when exposed with what the same group's outcome would have been if exposure had been absent. Consequently, the observed value of the outcome measure in the exposed group is compared with the expected value of the outcome measure that would have been observed in the exposed group if it had, hypothetically, not been exposed. The unexposed group's actual outcome is used as a proxy for the

Toward terminological, conceptual, and statistical clarity in the study of mediators and moderators: Examples from the child-clinical and pediatric psychology literatures.

Numerous recent attempts to identify mediated and moderated effects in child-clinical and pediatric research on child adjustment have been characterized by terminological, conceptual, and statistical inconsistencies. To promote greater clarity, the terms mediating and moderating are defined and differentiated. Recommended statistical strategies that can be used to test for these effects are reviewed (i.e., multiple regression and structural equation modeling techniques). The distinction between mediated and indirect effects is also discussed. Examples of troublesome and appropriate uses of these terms in the child-clinical and pediatric psychology literatures are highlighted.

Data reduction and statistical inconsistency in linear inverse problems

An estimator or confidence set is statistically consistent if, in a well-defined sense, it converges in probability to the truth as the number of data grows. We give sufficient conditions for it to be impossible to find consistent estimators or confidence sets in some linear inverse problems. Several common approaches to statistical inference in geophysical inverse problems use the set of models that satisfy the data within a χ 2 measure of misfit to construct confidence sets and estimates. For example, the minimum-norm estimate of the unknown model is the model of smallest norm among those that map into a χ 2 ball around the data. We give weaker conditions under which the χ 2 misfit approach yields inconsistent estimators and confidence sets. Both sets of conditions depend on a measure of the redundancy of the observations, with respect to an a priori constraint on the model. When the observations are sufficiently redundant, using a χ 2 measure of misfit to selected averages of the data yields consistent confidence sets and minimum-norm estimates. Under still weaker conditions, one can find consistent estimates and confidence intervals for finite collections of linear functionals of the model. In an idealization of the problem of estimating the Gauss coefficients of the magnetic field at the core from satellite data, using a constraint on the energy stored in the field, suitable data averaging leads to consistent confidence intervals for finite collections of the Gauss coefficients.

Foreign investment in Estonia: A statistical approach

STATISTICAL SURVEYS ON FOREIGN INVESTMENT in Estonia may present very different figures when compared to one another. The highest variation can be found in the total amount of foreign investment. Statistics published by the Estonian Ministry of Finance differ considerably from those published by the Estonian Central Statistical Office (see Appendix 1). Inconsistencies in statistics create a phenomenon in which mistakes snowball as figures pass from one clerk to another and finally end up in the hands of various interest groups who need statistics in their operations. This research aims first to locate the statistical pitfalls to which imperfect surveys often lead. Second, it studies the development of foreign investment in Estonia during the period from January 1987 to July 1994. The longitudinal analysis will help shed light on previously underexamined aspects of the statistics. This article is part of a research series investigating Finnish investment in the former Soviet Union. The first research on the business activities of Finnish companies in Soviet Estonia1 was carried out in 1991 (Liuhto, 1991). That study has since been followed by several others which have dealt with the development of entrepreneurship and foreign investment in Estonia and the northwest of Russia (Liuhto, 1992a; Liuhto, 1992b; Liuhto, 1994a; Liuhto, 1994b).

Ideology, Interest Group Scores, and Legislative Votes

This article reviews writings by economists and political scientists on the importance of ideology in governmental decision making. We consider various problems in this research, discuss several shortcomings in these recent studies in the context of the larger literature on legislative behavior, and demonstrate the statistical inconsistency of the commonly used methodologies based on interest group scores.

Inconsistencies in statistics of deaths from AIDS

From 1983 to 1988 there were 396 registrations of death certified as caused by immunodeficiencies including the acquired immunodeficiency syndrome (AIDS). Detailed examination of other death registration data yields an estimate of 692 deaths in this period from AIDS and its sequelae. Over the same period, 617 notified AIDS patients are known to have died. The data sources can be reasonably closely reconciled for the years 1983 to 1987, but there is an unexplained shortfall in reporting of deaths of notified AIDS patients in 1988. A study which matches notifications and death registrations for the same individuals is needed to explain these findings, which have implications for statistics of survival times in AIDS and for estimates of the size of the epidemic.

K X-ray emission cross sections by proton bombardment

X-ray emission cross sections by proton impact ionization of K-shell electrons are presented. These values are based upon weighted orthonormal fits to all the relevant experimental information available from the literature. Several methods for reducing possible statistical inconsistency among experimental data points are discussed in order to reach precise and accurate cross section estimates. Present results are compared to analytical cross section formulae and to variously fitted values.