FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W2060129191> ?p ?o ?g. }

• W2060129191 endingPage "33" @default.
• W2060129191 startingPage "27" @default.
• W2060129191 abstract "Nephrologists rely on valid clinical studies to inform their health care decisions. Knowledge of simple statistical principles equips the prudent nephrologist with the skills that allow him or her to critically evaluate clinical studies and to determine the validity of the results. Important in this process is knowing when certain statistical tests are used appropriately and if their application in interpreting research data will most likely lead to the most robust or valid conclusions. The research team bears the responsibility for determining the statistical analysis during the design phase of the study and subsequently for carrying out the appropriate analysis. This will ensure that bias is minimized and “valid” results are reported. We have summarized the important caveats and components in correctly choosing a statistical test with a series of tables. With this format, we wish to provide a tool for the nephrologist/researcher that he or she can use when required to decide if an appropriate statistical analysis plan was implemented for any particular study. We have included in these tables the types of statistical tests that might be used best for analysis of different types of comparisons on small and on larger patient samples. Nephrologists rely on valid clinical studies to inform their health care decisions. Knowledge of simple statistical principles equips the prudent nephrologist with the skills that allow him or her to critically evaluate clinical studies and to determine the validity of the results. Important in this process is knowing when certain statistical tests are used appropriately and if their application in interpreting research data will most likely lead to the most robust or valid conclusions. The research team bears the responsibility for determining the statistical analysis during the design phase of the study and subsequently for carrying out the appropriate analysis. This will ensure that bias is minimized and “valid” results are reported. We have summarized the important caveats and components in correctly choosing a statistical test with a series of tables. With this format, we wish to provide a tool for the nephrologist/researcher that he or she can use when required to decide if an appropriate statistical analysis plan was implemented for any particular study. We have included in these tables the types of statistical tests that might be used best for analysis of different types of comparisons on small and on larger patient samples. Clinical Summary•Analysis of data from a clinical study using incorrect statistical tests will likely result in invalid conclusions.•Practitioners of evidence-based nephrology should be able to recognize when the possibility exists that inappropriate statistical methods were used for data analysis in a particular study.•The choice of the proper statistical methods for the analysis of the data should occur during the study design phase and should be based on an understanding of the types and number of study variables and the anticipated study size.•Following a systematic approach that arises from an a priori identification of the outcomes and the types and scope of predictor or exposure variables to be measured in a clinical study will allow the informed clinician/investigator to determine the proper statistical methods for the most robust analysis. •Analysis of data from a clinical study using incorrect statistical tests will likely result in invalid conclusions.•Practitioners of evidence-based nephrology should be able to recognize when the possibility exists that inappropriate statistical methods were used for data analysis in a particular study.•The choice of the proper statistical methods for the analysis of the data should occur during the study design phase and should be based on an understanding of the types and number of study variables and the anticipated study size.•Following a systematic approach that arises from an a priori identification of the outcomes and the types and scope of predictor or exposure variables to be measured in a clinical study will allow the informed clinician/investigator to determine the proper statistical methods for the most robust analysis. In the other articles in this issue of Advances in Chronic Kidney Disease, the role of evidence in informing clinical care, the various study designs to address specific types of clinical questions, and the characteristics and limits of evidence that are unique to specific study designs will be described. An understanding of the sources of the bias inherent to a particular study design, of the degree of uncertainty imparted to the evidence by such biases, and of how one can minimize distortion of the truth that emerges from these biases is central to best practice of evidence-based medicine. An often under-appreciated issue distinct to a discussion of the rigors of study design is whether the authors used the best statistical tests to analyze their data. One cannot assume that the investigator’s choice of the optimal statistical test(s) is always rigorously vetted in the review process. Thus, a working knowledge of the proper use and limitations of statistical tests in the analysis of clinical trial data is an important foundation of evidence-based clinical practice. In this chapter, we will summarize in a series of tables designed as tools for the evidence-based medicine practitioner, the core statistical issues that, if met, will allow for the most unambiguous representation of the truth. Most readers assume that the journals ensure that authors have used the proper statistical tests for their data. This assumption rests on the belief that most journals require a statistician’s review of the statistical methods before publication. Although this might be current practice for a number of journals, it was less common in the past.1Omar R.Z. McNally N. Ambler G. et al.Quality research in healthcare: are researchers getting enough statistical support?.BMC Health Serv Res. 2006; 6: 2Crossref PubMed Scopus (5) Google Scholar, 2Goodman S.N. Altman D.G. George S.L. Statistical reviewing policies of medical journals. Caveat lector?.J Gen Intern Med. 1998; 13: 753-756Crossref PubMed Scopus (77) Google Scholar A survey of the editors of 171 high-impact medical journals in 1998 revealed that only one-third guaranteed statistical review for all accepted manuscripts.2Goodman S.N. Altman D.G. George S.L. Statistical reviewing policies of medical journals. Caveat lector?.J Gen Intern Med. 1998; 13: 753-756Crossref PubMed Scopus (77) Google Scholar More recently, editors continue to cite statistical problems as a major issue in submitted and published manuscripts,3Fernandes-Taylor S. Hyun J.K. Reeder R.N. et al.Common statistical and research design problems in manuscripts submitted to high-impact medical journals.BMC Res Notes. 2011; 4: 304Crossref PubMed Scopus (43) Google Scholar despite widespread use by authors of some statistical help.4Altman D.G. Goodman S.N. Schroter S. How statistical expertise is used in medical research.JAMA. 2002; 287: 2817-2820Crossref PubMed Scopus (98) Google Scholar It has been estimated that 37% to 60% of the published studies in the medical literature have used at least 1 incorrect statistical test.5Kanter M.H. Taylor J.R. Accuracy of statistical methods in TRANSFUSION: a review of articles from July/August 1992 through June 1993.Transfusion. 1994; 34: 697-701Crossref PubMed Scopus (40) Google Scholar, 6Gandhi R. Smith H.N. Mahomed N.N. et al.Incorrect use of the Student t test in randomized trials of bilateral hip and knee arthroplasty patients.J Arthoplasty. 2011; 26: 811-816Abstract Full Text Full Text PDF PubMed Scopus (6) Google Scholar Despite a growing awareness of the potential distortions of evidence that are introduced by incorrect study designs or statistical analyses,7Guyatt G.H. Sackett D.L. Cook D.J. Evidence-Based Medicine Working Group. Users’ guides to the medical literature: part II. How to use an article about therapy or prevention: part A. Are the results of the study valid?.JAMA. 1993; 270: 2598-2601Crossref PubMed Scopus (987) Google Scholar inappropriate statistical analysis does not seem to impact significantly the citation frequency of a particular manuscript.8Nieminen P. Carpenter J. Rucker G. et al.The relationship between quality research and citation frequency.BMC Med Res Methodol. 2006; 6: 42Crossref PubMed Scopus (116) Google Scholar The issue of incorrect choice of statistical tests remains significant for nephrology. Since the nephrology literature was globally appraised for quality of study design in 2004,9Strippoli G.F. Craig J.C. Shena F.P. The number, quality, and coverage of randomized controlled trials in nephrology.J Am Soc Nephrol. 2004; 15: 411-419Crossref PubMed Scopus (270) Google Scholar efforts to improve reporting of key components in clinical study design likely have resulted in some improvements in the published literature.10Hopewell S. Dutton S. Yu L.M. et al.The quality of reports of randomised trials in 2000 and 2006: comparative study of articles indexed in PubMed.BMJ. 2010; 340: c723Crossref PubMed Scopus (396) Google Scholar Nonetheless, a current review (December 9, 2011) of the instructions to authors/reviewers of 7 leading nephrology journals reveals that only 4 journals recommend that authors adhere to the Consolidated Standards of Reporting Trials (CONSORT) statement11David Moher D. Schulz K.F. Altman D. CONSORT GroupThe CONSORT Statement: revised recommendations for improving the quality of reports of parallel-group randomized trials.JAMA. 2001; 285: 1987-1991Crossref PubMed Scopus (2032) Google Scholar for the design and conduct of randomized clinical trials and only 1 journal endorses the comparable statements for other clinical study designs. None of these journals indicate in their instructions to authors or reviewers that a separate review by a statistician will be performed on all manuscripts accepted for publication. Paramount to maintaining the integrity of the data and validity of the reported study results is to select the most appropriate statistical analysis for the available variables and the type of question that the study is addressing. Herein, we present a tabular summary and guide to help the nephrology practitioner determine whether the correct or optimal statistical tests were selected by the author of published studies and his or her own selection when conducting original research. Current standard practice should have a statistician as part of the research team. Further, the design phase is when the statistical tests should be decided and planned, long before the data collection. Although it is best practice to have a statistician help design the study, advise the choice of statistical methods, and direct the conduct of the statistical analysis, a nephrologist skilled in statistical analysis will more likely design and conduct high-quality trails yielding meaningful results. Nephrologists armed with knowledge about proper use of statistical tests will more selectively apply the results of studies in the literature to improve their care of patients. Errors in proper statistical analysis can bias and skew the results, leading to incorrect conclusions that could potentially adversely impact clinical practice and patient care. Some common errors include the use of parametric tests on data that are nonparametric or likely skewed; applying a Student t test to a sample size that is insufficient; or misinterpreting a P value or depending on an inadequate level of significance or a study size, leading to a type I or type II error.12Strasak A.M. Zaman Q. Pfeiffer K.P. et al.Statistical errors in medical research–a review of common pitfalls.Swiss Med Wkly. 2007; 137: 44-49PubMed Google Scholar, 13Pocock S.J. Hughes M.D. Lee R.J. Statistical problems in the reporting of clinical trials. A survey of three medical journals.N Engl J Med. 1987; 317: 426-432Crossref PubMed Scopus (514) Google Scholar, 14Gardenier J.S. Resnik D.B. The misuse of statistics: concepts, tools, and a research agenda.Account Res. 2002; 9: 65-74Crossref PubMed Scopus (39) Google Scholar, 15McKinney W.P. Young M.J. Hartz A. et al.The inexact use of Fisher’s Exact Test in six major medical journals.JAMA. 1989; 261: 3430-3433Crossref PubMed Scopus (61) Google Scholar The primary objective of any clinical study must be to generate valid results while minimizing any opportunity for bias. It is otherwise unethical to expose subjects to the risks of participation in a clinical trial if the design, conduct, or analysis of the clinical trial is flawed. Reporting flawed results or conclusions potentially extends this harm to other patients. Producing valid and generalizable study results begins in the design phase with stating a clear clinical question to be addressed by the study, determining the variables to be collected, and determining the primary outcome to be measured. Knowing the exact question one plans to (or is able to) answer will ensure the collection of the correct types of data and the application of the appropriate statistical tests. All statistical analysis plans should be stated a priori, before the enrollment of the first study subject. Any post hoc analyses may be subject to investigator bias and yield results that at best can be used to generate new hypotheses that require further study for confirmation. Often, however, the “conclusions” that are generated by such post hoc analyses are not subsequently confirmed in properly designed trials. Determining which statistical tests are appropriate to a particular study depends first on understanding the types of outcomes data that will best address the central question of the study, the types and numbers of variables that will be collected to test the investigator’s hypothesis (typically of an association between an outcome and an intervention, exposure, clinical/demographic characteristic or risk factor), specific relationships of these variables to each other, and the size and complexity of data extracted from the sampled population. These most relevant issues determine what types of statistical tests one should expect to have been used by the investigator, and are summarized in the questions in Table 1.Table 1Questions to Help Determine the Correct Statistical Test1What type of outcomes measures one expects to find and how will these best be reported?2Which are the dependent and independent variables and what is the expected relationship between them?3How many independent variables are expected to impact the outcome (dependent variable(s)) under investigation?4Do the data allow for the use of parametric statistical tests or are nonparametric tests more appropriate? Open table in a new tab These questions apply to studies where the intervention or exposure is controlled by the investigator, as in a randomized controlled trial, that is, an experimental study, or where the intervention or exposure occurs naturally and is not under the control of the investigator, as in an epidemiologic or observational study. Table 2 lists each consideration and the subsequent related table defining in detail the relevant issues regarding test selection. Table 3, Table 4, Table 5, Table 6, Table 7, Table 8 describe the types of statistical tests used to interrogate data from various types of studies. Table 9 summarizes the general considerations for the diverse statistical analyses shown in Table 3, Table 4, Table 5, Table 6, Table 7, Table 8.Table 2Classification of Types of Studies and Scales/MeasuresConsideration About the DataScale/Type of MeasureTableOutcome/1 variable. Descriptive statisticsCentral tendency/dispersion3Comparison of 1 sample to populationRatio or interval, nominal4Comparison of 2 or more independent samples (1 independent/predictor variable)Ratio or interval, ordinal or discrete numerical, nominal5Comparison of 2 or more related (matched or paired) samples (1 independent/predictor variable)Ratio or interval, ordinal or discrete numerical, nominal6Measures of association (1 independent/predictor variable)Independent variableDependent variable7Nominal, ordinal, interval, or ratioNominal, ordinal, interval, or ratioTime to eventOrdinal (ranks)Internal/ratio (continuous/numerical)Measures of association (>1 independent/predictor variable [or confounder])See Table 88 Open table in a new tab Table 3Descriptive Statistics: 1 Variable/OutcomeScale/Type of MeasurementMeasureStatisticRatio or interval (continuous/numerical), normally distributedCentral tendencyMeanDispersionSDRatio or interval (continuous/numerical), not normally distributedCentral tendencyMedian, modeDispersionRange, IQROrdinal (ranks)Central tendencyMedian, modeDispersionRange, IQRNominal (categories)Central tendencyModeDispersionNumber of categoriesAbbreviations: SD, standard deviation; IQR, interquartile range. Open table in a new tab Table 4Comparison of 1 Sample to PopulationScale/Type of MeasurementRequirementTestRatio or interval (continuous/numerical)Normal distribution or n > 301-sample tn ≤ 30Wilcoxon signed rank testNominal (categories)1-sample binomialnp and n (1−p) >5Z approximation Open table in a new tab Table 5Comparison of 2 or More Independent Samples (Groups): 1 Independent/Predictor VariableScale/Type of MeasurementNumber of Samples (Groups)RequirementTestRatio or interval (continuous/numerical)2Normal distribution or n > 30, equal variancesT testNormal distribution or n > 30Welch’s T testRank sum test>2Normal distribution or n > 30, homogeneous varianceANOVAKruskal–WallisOrdinal (ranks) or discrete numerical2Rank sum test>2Kruskal–WallisNominal (categories)2Fisher exact test, χ2 test, other measures of association2Expected values ≥5χ2 testRelative risk (rate ratio)Odds ratio≥2Expected values ≥5χ2 testAbbreviation: ANOVA, analysis of variance. Open table in a new tab Table 6Comparison of 2 or More Related (Matched or Paired) Samples: 1 Independent/Predictor VariableScale/Type of MeasurementNumber of Samples (Measurements)RequirementTestRatio or interval (continuous/numerical)2Normal distribution or n > 30Paired t testWilcoxon signed rank test>2Normal distribution or n > 30, homogeneous varianceRepeated-measures ANOVAFriedman 2-way ANOVAOrdinal (ranks) or discrete numerical2Wilcoxon signed rank test>2Friedman 2-way ANOVANominal (categories)2McNemar χ2 test Open table in a new tab Table 7Measures of Association: 1 Independent/Predictor VariableScale/Type of MeasurementRequirementTestIndependent VariableDependent VariableNominal (categories)Nominal (categories), ordinal (ranks), interval/ratio (continuous/numerical)See comparison of 2 or more independent samples (groups) mentioned previouslyNominal (categories)Time to event (with or without censored observations)No secular trend. Outcome would be the same for those lost and followedKaplan–Meier survival analysisOrdinal (ranks)Ordinal (ranks)Spearman rank correlationInterval/ratio (continuous/numerical)Interval/ratio (continuous/numerical)Describes the magnitude and direction of numerical relationships. Joint distribution of x and y is normal. Linear (as opposed to nonlinear) relationshipPearson correlation coefficientPredicts the dependent variable from independent variable. Normal distribution and equal variance of y for any x. Linear (as opposed to nonlinear) relationshipSimple linear regression Open table in a new tab Table 8Measures of Association: >1 Independent/Predictor Variable or ConfounderScale/Type of MeasurementRequirementTestIndependent VariablesDependent VariableNominal (categories)Nominal (categories)Mantel–Haenszel (could also use logistic regression)Nominal (categories)Interval/ratio (continuous/numerical)Normal distribution or n > 30, homogeneous varianceFactorial ANOVANominal and interval/ratioInterval/ratio (continuous/numerical)Normal distribution or n > 30, homogeneous varianceANCOVANominal, ordinal, and/or interval/ratioInterval/ratio (continuous/numerical)Normal distribution and equal variance of y for any x and linear relationshipMultiple linear regressionNominal, ordinal, and/or interval/ratioNominal (categories), binary (only 2 categories)Linear relationship with odds for outcomeMultiple logistic regressionNominal, ordinal, and/or interval/ratioTime to event (with or without censored observations)Hazard ratio does not change over time, outcome the same for those lost and followedCox proportional hazard modelAbbreviation: ANCOVA, analysis of covariance. Open table in a new tab Table 9Summary of Most Appropriate Statistical Methods for Each Type of DataType of MeasurementType of DataParametricNonparametricNominalTime to EventDescriptive statistics of 1 groupMean, SDMedian, IQRProportion, modeKaplan–Meier survival plotCompare 1 sample to a population1-sample t testWilcoxon signed rank testZ approximation or binomial testKaplan–Meier survival plot or confidence intervalCompare 2 independent groupsUnpaired t testMann–Whitney test or rank sumFisher exact test or χ2 test (for larger samples)Log-rank test or Mantel–Haenszel testCompare 2 paired groupsPaired t testWilcoxon signed rank testMcNemar testConditional proportional hazards regressionCompare 3 or more independent groups1-way ANOVAKruskal–Wallis testχ2 test or Fisher exact testCox proportional hazard regressionCompare 3 or more matched groupsRepeated-measures ANOVAFriedman 2-way ANOVACochrane QConditional proportional hazards regressionMeasure of association between 2 variablesPearson correlationSpearman correlationContingency coefficientsKaplan–Meier survival plotMeasure of association of dependent variable with 1 independent variableSimple linear regressionNonparametric regressionSimple logistic regressionCox proportional hazard regressionMeasure of association of dependent variable with >1 independent variableMultiple linear regression or ANCOVANonparametric regressionMultiple logistic regressionCox proportional hazard regression Open table in a new tab Abbreviations: SD, standard deviation; IQR, interquartile range. Abbreviation: ANOVA, analysis of variance. Abbreviation: ANCOVA, analysis of covariance. Many outcomes of relevance to clinicians can be reported as numerical values or as ratios of numerical values that are continuous over a range of possible values, where each interval has discrete meaning. For example, the time after onset of nephrotic-range proteinuria to the development of ESRD in a patient with insulin-dependent diabetes mellitus and the duration and intensity of smoking before the development of lung cancer are typically reported as values along a continuum, as years or as pack-years, respectively. Other parameters or characteristics may form a clinically relevant hierarchy where the intervals between values do not have a discrete and consistent numerical meaning. Thus, patients with grade III hypertensive retinopathy have more severe disease than those with grade I, but the intervals between classes of hypertensive retinopathy do not indicate a specific magnitude of change. These ordinal values or ranks allow one to describe directionality of change but not the precise magnitude of change when, for instance, a patient progresses from grade I to grade II and then to grade III hypertensive retinopathy. Sometimes categories carry clinical meaning but do not describe a rank order. Thus, gender or ethnicity may be relevant clinical features for describing risk or modification of a response to an exposure but do not have an implicit rank order. These types of discrete, but nonranked, variables are called nominal data. Table 3 lists each of these types of measures or scales and the statistics that are used to describe their central tendency and their dispersion. The hypothesis under investigation may be as simple as determining whether a sample from a population differs in the feature of interest from the population as a whole, or as complex as determining whether multiple independent and dependent variables effect 2 or more subpopulations differently. When determining whether a specific sample of study subjects differs from the population as a whole, the analysis calls for the comparison of 1 sample to the population. As such analyses are not testing whether 1 factor results in an outcome, an a priori determination of the dependent and the independent variables is not mandated. These statistical tasks are shown in Table 4. In studies that are designed to test a hypothesis that 1 or more factors might result in the same or different outcomes in 1 or more populations, the correct statistical tests can be used only if there has been a correct hypothesis-driven assignment of the dependent and independent variables. The dependent variable is what the researcher is trying to predict, that is, the outcome measured. The hypothesis undergoing testing in a clinical trial then relates the effect the independent variable or variables exert on the dependent variable. Table 5, Table 6, Table 7, Table 8 summarize the statistical tests that are used to interrogate the relationships between the dependent and independent variables. The optimal statistical tests for the type of study question, the type and number of variables, and the study size are shown in Table 4, Table 5, Table 6, Table 7, Table 8. Table 4 was introduced previously for use when the analysis calls for the comparison of 1 sample to the population. Table 5 describes the statistical analyses used to compare 2 or more independent samples with 1 independent variable. For instance, one might wish to assess whether use of an angiotensin-converting enzyme inhibitor (ACEI) medication (independent variable) might lead to a reduced occurrence of ESRD in patients with diabetes mellitus. One might wish to similarly sample populations with different baseline demographic characteristics to determine whether ACEI use similarly reduced the development of ESRD among diabetic and nondiabetic patients, among the elderly versus younger patients, and among men versus women, or whether the effect of ACEIs on progression to ESRD among diabetics was influenced by ethnicity or race. Table 6 describes the statistical tests used to compare 2 or more matched or paired samples with 1 independent variable. Table 7, Table 8 describe the statistical tests used when evaluating associations, typically between an exposure or risk factor and an outcome for a single exposure (Table 7) or multiple independent exposures or predictors (Table 8). Table 9 highlights these processes and serves to summarize the important caveats to each test in a single table. A few additional special considerations will be addressed later in the text. A more detailed description of general statistical considerations for the informed evidence-based reader of the medical literature can be found in recent reviews of statistics.16Kennedy K. Frankowiski R.F. Evaluating the evidence about therapies: what the clinician needs to know about statistics.Clin Perinatol. 2003; 30: 205-215Abstract Full Text Full Text PDF PubMed Scopus (5) Google Scholar, 17McCrum-Gardner E. Which is the correct statistical test to use?.Br J Oral Maxillofacial Surg. 2008; 46: 38-41Abstract Full Text Full Text PDF PubMed Scopus (276) Google Scholar Selection of independent variables that are most likely to have some proximate relation to the outcome (or dependent variable) will reduce the risk of confounding. An a priori statement of variables to be measured and analyzed before the initiation of the study should conform to the major hypothesis under investigation. In less robustly designed studies, the investigator may believe it is prudent to collect all “possible” variables up front and then see which variables “fall out” in the analysis. Such an unsystematic approach to variable assignment, often termed a “fishing expedition” or “data mining,” can lead to false conclusions regarding apparent associations. Some limit of predictor variables per study participant is generally recommended, often no more than 5 to 10. When designing a study using existing data sets where variables have been predetermined, it remains important to state a priori the hypothesis under investigation and the variables that will be included in the analysis to test this hypothesis. It is important to ensure that the predictor variables are measured and entered into the analysis in the correct form (continuous or categorical). Each predictor variable should be assessed for its impact on the other variables. The impact of one variable on another may result from an interaction or an effect modification or may represent confounding. Whenever possible, the most important potential variables that may impact the outcome should be measured and included appropriately in the analysis. Failure to look for and identify significant confounders in the design and analysis phases of a study could lead to false conclusions. To use parametric statistical methods, the data must meet the assumptions that they are normally distributed and that sampling of the data is random. Parametric statistical tests appear in published papers where their use cannot be justified; most commonly, the use of parametric statistical methods on insufficiently sized samples to permit the assumption of a normal distribution. If these assumptions are not met, a nonparametric analysis should be performed. Table 4, Table 5, Table 6, Table 7, Table 8 differentiate when parametric and nonparametric tests should be used, and this is further summarized in Table 9. Nonparametric analysis may involve ranking the outcomes and analyzing the ranks. This is often used in situations where the outcome is a score or a rank value, where the data cannot assume a Gaussian distribution. If the data are continuous, but not normally distributed, they sometimes can be transformed so that they become normally distributed by taking the logarithm or inverse of the values, and thus in these circumstances, use of parametric tests is appropriate, even though the data before transformation were skewed. With a large sample size, parametric analysis is often appropriate even if the data are not perfectly normally distributed, but with small sample sizes, a normal or Gaussian distribution cannot be assumed. This approach is based on the central limit theorem that maintains that parametric testing is sufficiently robust to accommodate deviations from a Gaussian distribution as long as the sample size is sufficiently large (e.g., >30 subjects per group). Authors should state clearly the rationale for the choice of a specific statistical test and/or whether the assumptions underlying their choice of statistical methods were evaluated statistically. It is most frequently appropriate to use tests that generate a 2-sided P value. This ensures that the statistical methods used test for violation of the null hypothesis in both the negative and positive directions. If one chooses to use a 1-sided P value, the researcher must be certain that the difference in the outcomes can occur in 1 direction only. The likely 1-sidedness of the difference should be determined before any data collection and be most robustly biologically plausible. If it is not possible to determine a priori the directionality of the difference, a 2-sided P values should be used. A researcher’s “hunch” or biologic plausibility is not strong enough to justify a 1-sided P value, and therefore, 2-sided tests are typically more appropriate and are more conservative. A paired analysis is used when 2 groups are matched or 2 values are always linked. The statistical tests used for paired analysis are shown in Table 6. If, for example, one is evaluating the change in urinary protein excretion with or without an intervention, one would compare the value in the same patient in a paired manner. This type of analysis, when the data are linked, allows for a better correlation between the 2 sets of values or 2 groups than if each data point were compared with random values in the opposing group. Comparison of 3 or more groups is more involved statistically and may need to be addressed by a repeated-measures analysis. At this stage, all the independent and dependent variables have been categorized, the assumptions for each of the chosen statistical test reviewed, and the analysis most appropriate to the data determined. Any statistical test can be run on any set of data, but the results that such an analysis yields will likely be incorrect if statistical methods that are inappropriate for the types of data under investigation are used. As a responsible researcher and/or reviewer of the literature, possessing the ability to understand the basic tenets of statistical test selection will permit a more robust and judicious determination of the likely validity of a study’s reported results. Applying a few simple statistical principles allows for the selection of the appropriate statistical tests to perform the most credible and valid analysis with a given data set. The responsibility lies with the researcher to conduct and analyze the data honestly and with skill to ensure the integrity of the information reported. The responsibility of all practitioners of evidence-based nephrology is to critically appraise the published evidence to ensure that the conclusions reported with any study are based on the most robust analysis of the data using optimal methods to ensure the minimization of bias and of distortion of the “truth.” Thus, understanding the limits of and the correct application of statistical methods is a core skill of the evidence-based practitioner and researcher. The tables in the aforementioned discussion are intended as tools for the evidence-based nephrologist to allow them to ascertain quickly whether investigators likely have used the correct statistical methods in the analysis of their published data." @default.
• W2060129191 created "2016-06-24" @default.
• W2060129191 creator A5019781661 @default.
• W2060129191 creator A5044293139 @default.
• W2060129191 creator A5077422811 @default.
• W2060129191 date "2012-01-01" @default.
• W2060129191 modified "2023-09-26" @default.
• W2060129191 title "Evidence-Based Practice in Nephrology: Critical Appraisal of Nephrology Clinical Research: Were the Correct Statistical Tests Used?" @default.
• W2060129191 cites W1507860680 @default.
• W2060129191 cites W1746983276 @default.
• W2060129191 cites W1977833499 @default.
• W2060129191 cites W1992744347 @default.
• W2060129191 cites W1994930505 @default.
• W2060129191 cites W1995303893 @default.
• W2060129191 cites W2008946411 @default.
• W2060129191 cites W2010780893 @default.
• W2060129191 cites W2023284062 @default.
• W2060129191 cites W2041033959 @default.
• W2060129191 cites W2053958914 @default.
• W2060129191 cites W2070340893 @default.
• W2060129191 cites W2140262715 @default.
• W2060129191 cites W2142960568 @default.
• W2060129191 cites W2148195379 @default.
• W2060129191 cites W2152240392 @default.
• W2060129191 cites W2154317905 @default.
• W2060129191 cites W2156181132 @default.
• W2060129191 cites W36931139 @default.
• W2060129191 doi "https://doi.org/10.1053/j.ackd.2012.01.004" @default.
• W2060129191 hasPubMedId "https://pubmed.ncbi.nlm.nih.gov/22364798" @default.
• W2060129191 hasPublicationYear "2012" @default.
• W2060129191 type Work @default.
• W2060129191 sameAs 2060129191 @default.
• W2060129191 citedByCount "2" @default.
• W2060129191 countsByYear W20601291912015 @default.
• W2060129191 countsByYear W20601291912017 @default.
• W2060129191 crossrefType "journal-article" @default.
• W2060129191 hasAuthorship W2060129191A5019781661 @default.
• W2060129191 hasAuthorship W2060129191A5044293139 @default.
• W2060129191 hasAuthorship W2060129191A5077422811 @default.
• W2060129191 hasBestOaLocation W20601291911 @default.
• W2060129191 hasConcept C126322002 @default.
• W2060129191 hasConcept C142724271 @default.
• W2060129191 hasConcept C152541439 @default.
• W2060129191 hasConcept C177713679 @default.
• W2060129191 hasConcept C204787440 @default.
• W2060129191 hasConcept C2779974597 @default.
• W2060129191 hasConcept C512399662 @default.
• W2060129191 hasConcept C54847362 @default.
• W2060129191 hasConcept C71924100 @default.
• W2060129191 hasConceptScore W2060129191C126322002 @default.
• W2060129191 hasConceptScore W2060129191C142724271 @default.
• W2060129191 hasConceptScore W2060129191C152541439 @default.
• W2060129191 hasConceptScore W2060129191C177713679 @default.
• W2060129191 hasConceptScore W2060129191C204787440 @default.
• W2060129191 hasConceptScore W2060129191C2779974597 @default.
• W2060129191 hasConceptScore W2060129191C512399662 @default.
• W2060129191 hasConceptScore W2060129191C54847362 @default.
• W2060129191 hasConceptScore W2060129191C71924100 @default.
• W2060129191 hasIssue "1" @default.
• W2060129191 hasLocation W20601291911 @default.
• W2060129191 hasLocation W20601291912 @default.
• W2060129191 hasOpenAccess W2060129191 @default.
• W2060129191 hasPrimaryLocation W20601291911 @default.
• W2060129191 hasRelatedWork W135014142 @default.
• W2060129191 hasRelatedWork W2183606172 @default.
• W2060129191 hasRelatedWork W2783129026 @default.
• W2060129191 hasRelatedWork W2807375651 @default.
• W2060129191 hasRelatedWork W2884221656 @default.
• W2060129191 hasRelatedWork W2997026283 @default.
• W2060129191 hasRelatedWork W3011115316 @default.
• W2060129191 hasRelatedWork W4214655940 @default.
• W2060129191 hasRelatedWork W4252371801 @default.
• W2060129191 hasRelatedWork W21757 @default.
• W2060129191 hasVolume "19" @default.
• W2060129191 isParatext "false" @default.
• W2060129191 isRetracted "false" @default.
• W2060129191 magId "2060129191" @default.
• W2060129191 workType "article" @default.