Nonrandom Errors Research Articles

A-140 Reducing cell count bias with the GloCyte® Automated Cell Counter for CSF: an observational study at Loyola University Medical Center

Abstract Background Accurate STAT cerebrospinal fluid (CSF) cell counts are critical in supporting diagnoses of meningitis, malignancy, demyelinating disease, and hemorrhage. Limitations of automated analyzers to achieve acceptable linearity for the clinically relevant range has maintained manual CSF cell counting as the preferred method. Manual cell counting, performed with a hemocytometer and bright field microscope, is time-consuming, requires specialized skills and depends on human interpretation. This subjectivity inherently introduces the potential for non-random error in the enumeration of both total nucleated cells (TNC) and red blood cells in CSF by manual methods. The purpose of this study was to evaluate and compare the level of bias inherent in manual determination (TNC) in CSF as compared to the GloCyte Automated Cell Counter. Methods Result bias data for CSF specimens processed in the hematology lab at Loyola University Medical Center, a quaternary care 547-bed academic medical center in Maywood, IL, was analyzed for manual cell counting using a Neubauer-improved hemocytometer and GloCyte, an FDA-cleared cell counting device employing fluorescence technology to automate RBC and TNC counts in under 5 minutes, with linearity down to 0 cells/µL. Result bias was determined by defining the “expected result” for both GloCyte and hemocytometer as the initial cell count determined at a 1:5 dilution of a freshly-collected CSF sample and calculating the expected cell count for subsequent serial dilutions of 1:10, 1:20, 1:50, 1:100 for each method. The actual TNC count and the disparity of the actual result from the expected result was recorded as bias. Results Data are summarized below in the Table, and indicate a lower bias for GloCyte compared to manual counts across the range of dilutions. Conclusions This study demonstrates that GloCyte can be implemented as a means of effectively mitigating the reporting bias of CSF TNC counts performed with a Neubauer-improved hemocytometer.

Foveal crowding for large and small Landolt Cs: Similarity and Attention

We compare the recognition of foveal crowded Landolt Cs of two sizes: brief (40 ms), large, low-contrast Cs and high-contrast (1 sec) tests at the resolution limit of the visual system. In different series, the test Landolt C was surrounded by two identical distractors located symmetrically along the horizontal or by a single distractor. The distractors were Landolt Cs or rings. At the resolution limit, the critical spacing was similar in the two series and did not depend on the type of distractor. The result supports the hypothesis that crowding at the resolution limit occurs when both the test and the distractors fall into the same smallest receptive field responsible for the target recognition. For large stimuli, at almost all separations distractors of the same shape caused greater impairment than did rings, and recognition errors were non-random. The critical spacing was equal to 0.5 test diameters only in the presence of one distracting Landolt C. This result suggests that attention is involved: When one distractor is added, involuntary attention, which is directed to the centre of gravity of the stimulus, can lead to confusion of features that are present in both tests and distractors and thus to non-random errors.

Which robust regression technique is appropriate under violated assumptions? A simulation study

Ordinary least squares (OLS) regression is widely employed for statistical prediction and theoretical explanation in psychology studies. However, OLS regression has a critical drawback: it becomes less accurate in the presence of outliers and non-random error distribution. Several robust regression methods have been proposed as alternatives. However, each robust regression has its own strengths and limitations. Consequently, researchers are often at a loss as to which robust regression method to use for their studies. This study uses a Monte Carlo experiment to compare different types of robust regression methods with OLS regression based on relative efficiency (RE), bias, root mean squared error (RMSE), Type 1 error, power, coverage probability of the 95% confidence intervals (CIs), and the width of the CIs. The results show that, with sufficient samples per predictor (n = 100), the robust regression methods are as efficient as OLS regression. When errors follow non-normal distributions, i.e., mixed-normal, symmetric and heavy-tailed (SH), asymmetric and relatively light-tailed (AL), asymmetric and heavy-tailed (AH), and heteroscedastic, the robust method (GM-estimation) seems to consistently outperform OLS regression.

Comparison between intraoral scanning and direct visual analysis for the assessment of developmental defects of enamel

ObjectiveTo compare direct visual analysis (DVA) and intraoral scanning (IOS) for the assessment of developmental defects of the enamel (DDE). MethodsThirty-nine extracted permanent human teeth with DDE were selected by an experienced examiner and digitised using IOS. The scanning was recorded using the OBS Studio software parallel to the IOS software to obtain a coloured high-definition MP4 file of the process. Two other experienced, blinded, and calibrated examiners randomly analysed the same teeth through DVA and IOS. A third examiner resolved any disagreements between the two examiners. Descriptive statistics were used to analyse the frequencies of the scores. Cohen's kappa test was used to determine whether the DVA scores were different from those assigned using IOS. Spearman's test was used to verify non-random examiner errors. The Chi-square test was used to compare score frequencies. Statistical significance was set at p <0.05. ResultsScores indicating more severe and extended DDE (p <0.05) were more frequently assigned with IOS than with DVA (IOS: 25.64%, 25.64%, 38.46%, and 35.90% between one-third to two-third of the lingual, occlusal, mesial, and distal surfaces, respectively; vs. DVA: 10.26%, 7.69%, 15.38%, and 10.26% for the respective aforementioned tooth surfaces). Contrarily, ‘no visible enamel defect’ was significantly less assigned for IOS than for DVA (IOS: 15.38%, 43.59%, 35.90%, 15.38%, and 17.95% for buccal, lingual, occlusal, mesial, and distal surfaces, respectively; vs. DVA: 38.46%, 66.67%, 56.41%, 51.28%, and 43.59% for the respective aforementioned tooth surfaces). Kappa agreement ranged from fair to moderate when comparing DVA and IOS; the correlation between both methods was positive, indicating that the examiners assigned the scores properly and the differences arose from employing different methods. ConclusionThe assessment of DDE differed depending on the method used. IOS scores indicated more severe and extended DDE than DVA scores. Clinical investigation is the next step in validating the use of IOS for DDE diagnosis. Clinical SignificanceThis study showed that DDE can be assessed differently using IOS. It is clinically relevant as it directly affects the determination of the severity of the defect and dental treatment planning.

Mapinsights: deep exploration of quality issues and error profiles in high-throughput sequence data.

High-throughput sequencing (HTS) has revolutionized science by enabling super-fast detection of genomic variants at base-pair resolution. Consequently, it poses the challenging problem of identification of technical artifacts, i.e. hidden non-random error patterns. Understanding the properties of sequencing artifacts holds the key in separating true variants from false positives. Here, we develop Mapinsights, a toolkit that performs quality control (QC) analysis of sequence alignment files, capable of detecting outliers based on sequencing artifacts of HTS data at a deeper resolution compared with existing methods. Mapinsights performs a cluster analysis based on novel and existing QC features derived from the sequence alignment for outlier detection. We applied Mapinsights on community standard open-source datasets and identified various quality issues including technical errors related to sequencing cycles, sequencing chemistry, sequencing libraries and across various orthogonal sequencing platforms. Mapinsights also enables identification of anomalies related to sequencing depth. A logistic regression-based model built on the features of Mapinsights shows high accuracy in detecting 'low-confidence' variant sites. Quantitative estimates and probabilistic arguments provided by Mapinsights can be utilized in identifying errors, bias and outlier samples, and also aid in improving the authenticity of variant calls.

Measuring the impact of tax loss carryforwards on capital structure choice: the accuracy of earnings-based proxies

Empirical studies on capital structure choice frequently control for whether or not a firm has tax loss carryforwards (TLCFs). However, tax data on TLCFs is often not available to researchers. As a consequence, many studies rely upon earnings-based information in order to determine whether a firm has TLCFs. In this paper, we study the accuracy of earnings-based proxies in predicting whether a firm has TLCFs. Our results, which are based upon a panel of Italian listed firms between 2009 and 2012, show that earnings-based proxies correctly predict whether a firm has TLCFs for 70–77% of all firm-year observations. Furthermore, for all but one proxy, we identify a non-random measurement error in TLCF variables that are built upon these proxies. We evaluate the consequences of this measurement error and find that it does not significantly change the estimated effects of TLCFs on capital structure choice for some proxies, including last year’s earnings before taxes. We conclude that last year’s earnings before taxes serves as an accurate proxy for a firm’s true TLCF status. It is easy to determine and leads to statistically and quantitatively similar conclusions as those reached using tax data on TLCFs.

A comparative study of anomaly detection methods for gross error detection problems

The chemical industry requires highly accurate and reliable measurements to ensure smooth operation and effective monitoring of processing facilities. However, measured data inevitably contains errors from various sources. Traditionally in flow systems, data reconciliation through mass balancing is applied to reduce error by estimating balanced flows. However, this approach can only handle random errors. For non-random errors (called gross errors, GEs) which are caused by measurement bias, instrument failures, or process leaks, among others, this approach would return incorrect results. In recent years, many gross error detection (GED) methods have been proposed by the research community. It is recognised that the basic principle of GED is a special case of the detection of outliers (or anomalies) in data analytics. With the developments of Machine Learning (ML) research, patterns in the data can be discovered to provide effective detection of anomalous instances. In this paper, we present a comprehensive study of the application of ML-based Anomaly Detection methods (ADMs) in the GED context on a number of synthetic datasets and compare the results with several established GED approaches. We also perform data transformation on the measurement data and compare its associated results to the original results, as well as investigate the effects of training size on the detection performance. One class Support Vector Machine outperformed other ADMs and five selected statistical tests for GED on Accuracy, F1 Score, and Overall Power while Interquartile Range (IQR) method obtained the best selectivity outcome among the top 6 AMDs and the five statistical tests. The results indicate that ADMs can potentially be applied to GED problems.

Chromosome Inequality: Causes and Consequences of Non-Random Segregation Errors in Mitosis and Meiosis.

Aneuploidy is a hallmark of cancer and a major cause of miscarriages in humans. It is caused by chromosome segregation errors during cell divisions. Evidence is mounting that the probability of specific chromosomes undergoing a segregation error is non-random. In other words, some chromosomes have a higher chance of contributing to aneuploid karyotypes than others. This could have important implications for the origins of recurrent aneuploidy patterns in cancer and developing embryos. Here, we review recent progress in understanding the prevalence and causes of non-random chromosome segregation errors in mammalian mitosis and meiosis. We evaluate its potential impact on cancer and human reproduction and discuss possible research avenues.

Opportunities and illusions of using large samples in statistical inference

The theory of statistical inference clearly describes the benefits of large samples. The larger the sample size, the fewer standard errors of the estimated population parameters (the precision of the estimation improves) and the values of the power of statistical tests in hypothesis testing increase. Today’s easy access not only to large samples (e.g. web panels) but also to more advanced and user-friendly statistical software may obscure the potential threats faced by statistical inference based on large samples. Some researchers seem to be under the illusion that large samples can reduce both random errors, typical for any sampling technique, as well as non-random errors. Additionally, the role of a large sample size is an important aspect of the much discussed in the recent years issue of statistical significance (p-value) and the problems related to its determination and interpretation. The aim of the paper is to present and discuss the consequences of focusing solely on the advantages of large samples and ignoring any threats and challenges they pose to statistical inference. The study shows that a large-size sample collected using one of the non-random sampling techniques cannot be an alternative to random sampling. This particularly applies to online panels of volunteers willing to participate in a survey. The paper also shows that the sampling error may contain a non-random component which should not be regarded as a function of the sample size. As for the contemporary challenges related to testing hypotheses, the study discusses and exemplifies the scientific and ethical aspects of searching for statistical significance using large samples or multiple sampling.

Optical and structural optimization of a large aperture solar parabolic trough collector

Concentrating solar power is well established renewable energy technology. It involves the same heat engine cycle as the conventional power plants, and the heat produced can be stored effectively. However, expensive mirrors and structural support systems required to maintain high precision in the shape of reflectors for better optical and thermal efficiencies are the challenges involved in making it competitive with other solar power technologies. In the present work, a methodology is proposed based on the finite element method and Monte-Carlo ray tracing to design and optimize a parabolic trough collector. The effect of geometric and optical parameters such as rim angle, the width of the collector, tracking errors, and slope errors are analyzed. The effect of tracking errors in the East-West direction and slope errors are substantial compared to the other non-random errors. Loss of 25% and 7.1% in optical efficiency occurs for tracking errors of 20 mrad and slope error of 3.2 mrad, respectively. Based on the structural and optical study, an optimization method and optimum configurations are proposed. Using the proposed methodology, heat flux maps at the receiver can also be obtained for the structural health monitoring of the receiver tube.

What is Collider Bias and Why Should We Care?

Bias can threaten any study's internal validity compromising the findings. Bias is a nonrandom error that may affect the relationship between an exposure, risk factor, or treatment and the outcome. Bias can arise at any point in the research process, be it in the early stages of sample selection or the final stages of result rationalization and publication. There are 3 main types of bias: selection bias, information bias, and confounding bias.1

Machine-learning based prediction of crash response of tubular structures

This paper proposes a machine learning based methodology for predicting the buckling response of tubular structures. An extensive dataset of force-time curves is generated using a calibrated finite element model within a parametric space where buckling response is highly non-linear. Based on a fully connected neural network template, the machine learning hyper-parameters are determined and the resulting model is evaluated on a separate test set, with regard to maximum and average load and energy absorption errors. This evaluation shows a non-random error distribution which can be correlated with the physical properties of the structural collapse. To validate this assumption, a similar error analysis is conducted between finite element simulations with varying geometric imperfections. Evaluation of imperfection sensitivity reveals a similar error distribution and comparison of individual curves shows that errors made by the neural network model have a physical interpretation. These results indicate that the proposed machine learning based approach is capable of predicting the crushing response with a level of accuracy comparable to the errors that would be caused by a minor change in geometric imperfection.

Nano2NGS-Muta: a framework for converting nanopore sequencing data to NGS-liked sequencing data for hotspot mutation detection.

Nanopore sequencing, also known as single-molecule real-time sequencing, is a third/fourth generation sequencing technology that enables deciphering single DNA/RNA molecules without the polymerase chain reaction. Although nanopore sequencing has made significant progress in scientific research and clinical practice, its application has been limited compared with next-generation sequencing (NGS) due to specific design principle and data characteristics, especially in hotspot mutation detection. Therefore, we developed Nano2NGS-Muta as a data analysis framework for hotspot mutation detection based on long reads from nanopore sequencing. Nano2NGS-Muta is characterized by applying nanopore sequencing data to NGS-liked data analysis pipelines. Long reads can be converted into short reads and then processed through existing NGS analysis pipelines in combination with statistical methods for hotspot mutation detection. Nano2NGS-Muta not only effectively avoids false positive/negative results caused by non-random errors and unexpected insertions-deletions (indels) of nanopore sequencing data, improves the detection accuracy of hotspot mutations compared to conventional nanopore sequencing data analysis algorithms but also breaks the barriers of data analysis methods between short-read sequencing and long-read sequencing. We hope Nano2NGS-Muta can serves as a reference method for nanopore sequencing data and promotes higher application scope of nanopore sequencing technology in scientific research and clinical practice.

The difficulty of making claims to knowledge in social science

This paper considers three common types of claim to research knowledge, and the relative difficulty of making each type of claim in an empirically and logically justified manner. Before this, the paper looks at some more general issues often raised when discussing knowledge in social science, such as the nature of truth and justified belief, the existence of “isms” and paradigms treated like fashion accessories that one can adopt or not at will, and the intrinsic limitations of how we get to know about the “stuff” we might want to make research claims about. The idea of this early section is to remove some potential obstacles, before arguing that none of these issues is relevant to the rest of the paper about the nature of claims in their most generic form, independent of things like specific methods of data collection.&#x0D; &#x0D; The first type of claim we identify is a fully descriptive one that only summarises the data observed. This is the easiest and safest kind of claim, but even these might suffer from non-random errors and inaccuracies. However, their biggest limitation is their lack of any wider utility. The second kind of claim is a generally descriptive one that makes statements about unobserved data on the basis of a fully descriptive claim. Here we meet Hume’s problem of induction. These claims have two parts, and the inductive part cannot seemingly be justified by logic, inferential statistics (whether Fisher or Neyman-Pearson style), Carnap’s inductive probabilities, or even necessarily by Popper’s falsification process. The third type is a causal claim, which we argue must also be a general claim. We develop a model, based on the work of Mill and Bradford-Hill, of what a plausible causal claim entails. But it still has all of the problems emerging from the first two types of claim, and adds a further problem created by our inability to assess causes directly. The paper concludes by suggesting how social science can proceed most safely in practice.&#x0D;

Reviewing the logic of social scientific claims

This paper considers three different claims to knowledge, namely, “fully descriptive”, “generally descriptive” and causal claims. These are all common in social science, and each type of claim requires more assumptions than the previous one. After discussing their methodological and logical foundations, this paper describes some of the limitations in the nature of these three claims. Fully descriptive claims suffer from non-random errors and inaccuracies in observations, and can be queried in terms of utility. Generally, in addition to observational errors, descriptive can be questioned because of the long-standing problem of induction. Even the notion of falsification might not be able to help with this. Finally, causal claims are the most problematic of the three. While widely assumed, causation cannot be observed directly. The paper combines and develops three models of what causation might be, and discusses their implications for causal claims. It points out that so far our belief in causation is still a kind of religious one, and that neither theory nor inferential statistics can help in proving or observing its existence. Finally, the paper provides some suggestions for avoiding being misled by false knowledge and reporting our research findings with tentative care and judgement.

Quantitative Bias Analysis of the Association between Occupational Radiation Exposure and Ischemic Heart Disease Mortality in UK Nuclear Workers.

The scientific question of whether protracted low-dose or low-dose-rate exposure to external radiation is causally related to the risk of circulatory disease continues to be an important issue for radiation protection. Previous analyses of a matched case-control dataset nested in a large cohort of UK nuclear fuel cycle workers indicated that there was little evidence that observed associations between external radiation dose and ischemic heart disease (IHD) mortality risk [OR = 1.35 (95% CI: 0.99-1.84) for 15-year-lagged exposure] could alternatively be explained by confounding from pre-employment tobacco smoking, BMI or blood pressure, or from socioeconomic status or occupational exposure to excessive noise or shiftwork. To improve causal inference about the observed external radiation dose and IHD mortality association, we estimated the potential magnitude and direction of non-random errors, incorporated sensitivity analyses and simulated bias effects under plausible scenarios. We conducted quantitative bias analyses of plausible scenarios based on 1,000 Monte Carlo samples to explore the impact of exposure measurement error, missing information on tobacco smoking, and unmeasured confounding, and assessed whether observed associations were reliant on the inclusion of specific matched pairs using bootstrapping with 10% of matched pairs randomly excluded in 1,000 samples. We further explored the plausibility that having been monitored for internal exposure, which was an important confounding factor in the case-control analysis for which models were adjusted, was indeed a confounding factor or whether it might have been the result of some form of selection bias. Consistent with the broader epidemiological evidence-base, these analyses provide further evidence that the dose-response association between cumulative external radiation exposure and IHD mortality is non-linear in that it has a linear shape plateauing at an excess risk of 43% (95% CI: 7-92%) on reaching 390 mSv. Analyses of plausible scenarios of patterns of missing data for tobacco smoking at start of employment indicated that this resulted in relatively little bias towards the null in the original analysis. An unmeasured confounder would have had to have been highly correlated (rp > 0.60) with cumulative external radiation dose to importantly bias observed associations. The confounding effect of "having been monitored for internal dose" was unlikely to have been a true confounder in a biological sense, but instead may have been some unknown factor related to differences over time and between sites in selection criteria for internal monitoring, possibly resulting in collider bias. Plausible patterns of exposure measurement error negatively biased associations regardless of the modeled scenario, but did not importantly change the shape of the observed dose-response associations. These analyses provide additional support for the hypothesis that the observed association between external radiation exposure and IHD mortality may be causal.

Full-scale production of high-quality wood pellets assisted by multivariate statistical process control

In order for solid biofuels to be marketable and accessible to stakeholders (e.g., manufacturers, transhippers and customers), they must strictly meet the rigors of international standards and regulatory agencies regarding quality. On this basis, this study assesses the production of high-performance fuel-flexible woody pellets for heating and power, with the assistance of multivariate statistical process control (MSPC). The operational scenarios of industrially pelletizing pinewood sawdust are POSI (150 MPa and no pre-heating) and POSII (250 MPa and no pre-heating) as references, and POSIII (150 MPa and pre-heating) and POSIV (250 MPa and pre-heating) as alternatives to regularize the workflow on the machine. The stream-monitoring techniques of MSPC are the exponential moving average (EMA) and Hotelling's T2. The EMA-Hotelling system accurately tracks non-random errors and predicts the operationally finest scenario for developing high-quality pellets. Pre-heating the compressing channel of the machine regularizes the feeding and thus produces type I pellets (class A1/A2, European standard) with excellent bulk density (1227.55 kg m−3), durability (98.10%) and hydrophobicity (96.65%), without points outside the specific critical ranges. Clear evidence of improved pellet quality with the assistance of MSPC is found. The high-fidelity concept of MSPC captures the advantages of EMA and Hotelling's T2 into an immersive single framework; hence, it can compensate for the potential bias and misinformation-to-information overlapping of outcomes from monitoring latently multivariable flow on classical quality control charts. The major findings and innovations of this study can assist with rolling-out fine-scale pelletization and thus fulfill the increasing global demand for high-quality biofuels.

Automatic Methods for Detecting Cultural Bias in Social Media (Based on Telegram’s Dialogs)

In this paper, we described and tested several ways to use machine learning in order to analyze large collections of text data from social networks (namely, public Telegram chat), retrieve relevant social or cultural information from them, and to visualize the results of the research. The proposed approach has an advantage to reveal hidden patterns of social, political or cultural behavior by being able to cover large amounts of data. It can complement the standard social surveys methodology. Automatic detecting cultural bias on the example of social media requires mastering methods for measuring and visualizing its different kinds, such as cultural shifts, specific national or group refractions, mutations, stereotypes. We argue that cultural bias is a result of nonrandom errors in thinking. It is based, firstly, on a person's understanding of himself and the world around him and, secondly, on the translation of this understanding into abstraction in the form of common misconceptions, ideologemes, narrative, slogans. In society the bias inevitably leads to the separation of one social group or subculture from another. Social networks (both classic and new formats, for example, messengers with public chat options) are the most active ground for the representation of this phenomenon. Since the discussion of sociopolitical and cultural contexts in the case of chats takes place in public, the participants of such a communicative act tend to get approval of the social group to which they are ideologically close. It is this phenomenon that allows us to form comparisons of the “friend - foe” type, which lead next to unconscious cultural shifts. Thus, mastering methods to identify properly cultural shifts is not only relevant but crucial for the intra- and intercultural communication, for controlling the level of aggressiveness of the society, understanding its mood. As helpful illustrations, readers will find semantic associations elicited by the words “freedom”, “democracy”, “Internet”; sociocultural analysis of several topical clusters (e.g. Россия, страна, Путин, русский, православный); visualization of semantic associations for the words “freedom”, “democracy”, “Internet”.

Accuracy and uncertainty analysis of staple food crop modelling by the process-based Agro-C model.

Accuracy analysis of a process-based model is important for evaluating the reliability of model estimates of crop growth. Uncertainties in projections of crop growth may derive from different sources in modelling. The parameter-induced uncertainty is one of the important aspects. Here we calibrated the parameters for rice, wheat and maize combined with observed data of aboveground biomass (AGB) and leaf area index (LAI) at 16 Chinese Ecosystem Research Network (CERN) sites under different rotation systems and subsequently validated the model at these sites using the data independent of calibration. The results showed that the simulated AGB and LAI exhibited good agreement with the observations. The model performance for rice and maize was better than that for wheat. The statistical analysis of model performance showed that the RMSE (root mean square error), RMD (relative mean deviation) and EF (model efficiency) were 32.52%, - 0.95% and 0.87 of the means, respectively. The three components of the modelling uncertainty, bias of mean (UM), bias of slope (UR) and random residue (UE) accounted 0.1%, 0.9% and 99% of the total errors, respectively. The main contributor to the error was the random disturbances, indicating that the parameters calibration in this study had reached relatively reasonable conditions on the whole. Although the model displayed an overall good prediction in crops AGBs and LAI, there were still notable bias at some sites due to non-random errors (UM and UR). This indicated that there were still uncertainties in the modelling procedure, e.g. the model mechanism or parameterization. The uncertainty of the simulated results may greatly restrict the application of a model. To effectively and reasonably apply a model, it is necessary to evaluate and analyse the main sources of uncertainty in the simulated results. The parameter-induced uncertainty analysis in this study showed that, at the site scale, the range of uncertainty brought by the changes in three parameters (SLA, PL and α) to the modelling results (95% CI) of Agro-C covered more than 90% of the observations and brought approximately 21% uncertainty to the simulated AGBs of the three major crops.

Non-random errors and their importance in testing of hypotheses

Increasing numbers of non-random errors are observed in contemporary sample surveying – in particular, those resulting from no response or faulty measutrements (imprecise statistical observation). Until recently, the consequences of these kinds of errors have not been widely discussed in the context of the testing of hypoteses. Researchers focused almost entirely on sampling errors (random errors), whose magnitude decreases as the size of the random sample grows. In consequence, researchers who often use samples of very large sizes tend to overlook the influence random and non-random errors have on the results of their study. The aim of this paper is to present how non-random errors can affect the decision-making process based on the classical hypothesis testing procedure. Particular attention is devoted to cases in which researchers manage samples of large sizes. The study proved the thesis that samples of large sizes cause statistical tests to be more sensitive to non-random errors. Systematic errors, as a special case of non-random errors, increase the probability of making the wrong decision to reject a true hypothesis as the sample size grows. Supplementing the testing of hypotheses with the analysis of confidence intervals may in this context provide substantive support for the researcher in drawing accurate inferences.