# Research and Statistics Dictionary

## Definitions for a wide spectrum of empirical and statistical constructs

### A

**- ARI is the absolute difference between the control event rate (CER) and the experimental event rate (EER), |CER-EER|. ARI is necessary for calculating number needed to harm (NNH). The formula for NNH is (1/ARI).**

__Absolute risk increase (ARI)__**- ARR is the absolute difference between the experimental event rate (EER) and the control event rate (CER), |EER-CER|. ARR is necessary for calculating number needed to treat (NNT). The formula for NNT is (1/ARR).**

__Absolute risk reduction (ARR)____- Accuracy in measurement relates the validity, interpretability, and utility of a variable or outcome.__

**Accuracy in measurement**__- When testing multiple comparisons concurrently, the chances of committing a Type I error by chance alone increase drastically (known as experimentwise error rates). There are several methods for combating these errors: Bonferroni, Tukey's HSD, and Scheffe's test.__

**Adjusting for multiple comparisons**__- The alpha value is the criterion set by researchers that assumes statistical significance as well as the chance they are willing to take for committing a Type I error.__

**Alpha Level****- A type of reliability evidence where two forms of a survey instrument are written to cover the exact same content areas. The instruments should be highly correlated if alternate/parallel forms reliability is assumed.**

__Alternate/Parallel forms reliability__**- ANCOVA is a between-subjects statistical test that adjusts the outcome variable when comparing three or more independent groups.**

__Analysis of covariance (ANCOVA)__**- ANOVA is a between-subjects statistical test used when comparing three or more independent groups on a continuous outcome. The statistical assumptions of independence of observations, normality, and homogeneity of variance must be met in order to conduct ANOVA. If a significant main effect is found,**

__Analysis of variance (ANOVA)__*p*< .05, then post hoc tests must be used to explain the main effect.

### B

**- A within-subjects measure of baseline frequency used as the control.**

__Baseline Frequency__**- A within-subjects measure of baseline median values used as the control.**

__Baseline Median__**- A within-subject measure of baseline mean values used as the control.**

__Baseline Observation____- A statistical method used to predict the probability of an outcome based on the pretest probability of the outcome based on clinical factors and 2) the ability of diagnostic tests.__

**Bayes' Theorem**__- The beta value is used to denote how much of a chance a researcher is willing to take for committing a Type II error. The beta value is also used to calculate statistical power (1 - beta value = statistical power).__

**Beta Level****- A statistical method for comparing independent groups on an outcome.**

__Between-subjects__**- A statistical test of magnitude and direction of association between an ordinal variable and a continuous variable.**

__Biserial__**- A process used in experimental designs, such as randomized controlled trials (RCT), whereby study participants, clinicians, and/or researchers are not aware if a treatment or a placebo is being administered. Blinding is used to deter observation biases in experimental research. A single-blinded study is when study participants are now aware if they are being given the treatment or the placebo. A double-blinded study is when study participants and clinicians do not know who is receiving the treatment or placebo. Triple-blinded studies are the most rigorous in that the study participants, clinicians, and the researchers are unaware of the treatment or placebo and outcomes are assessed in an independent fashion.**

__Blinding____- A method of randomization where study participants are randomly allocated to groups in small blocks of four to six "blocks" at a time. This method ensures equally sized groups but is only feasible in smaller studies.__

**Blocked randomization****- A framework for understanding different levels of "knowing" related to a content area or performance.**

__Bloom's Taxonomy____- A post hoc test used when testing multiple hypotheses at the same time. The test is often used to correct for familiywise error rates (increased Type I error rates). Take the alpha value being used, often .05, and divide it by the number of statistical tests being conducted. That value is the new alpha value used to assume statistical significance. For example, you are testing eight hypotheses with an alpha of .05, .05/8 = .006. The p-values for each test will have to now be below .006 to achieve statistical significance.__

**Bonferroni****- A method for validating statistical findings. Thousands of random samples are taken in a non-exclusive fashion and statistics are rerun on each sample. The method yields 95% confidence intervals for statistical findings.**

__Bootstrap__### C

**- A type of observational research design that establishes associations between outcome variables and potential predictor variables. It is considered a weaker type of observational design due to the amount of bias caused by retrospectively selecting cases and controls and analyzing data. Case-control designs are well-suited for studying rare outcomes and generating hypotheses. Odds ratios are used to establish associations in case-control designs.**

__Case-control__**- A type of observational research design that yields the lowest level of empirical evidence. A series of observations are taken from a given population in a retrospective fashion and analyzed. Case series designs are useful for studying rare outcomes and hypothesis generation.**

__Case series__**- A type of variable that denotes group membership, possession of a characteristic or trait, or categorization of phenomena using numerical values.**

__Categorical variable__**- A between-subjects statistical test where two independent groups are compared on a dichotomous categorical outcome.**

__Chi-square__**- A statistical test where the expected dispersal of proportions in levels of a categorical variable is compared to the observed dispersal of proportions in levels of a a categorical variable.**

__Chi-square Goodness-of-fit__**- A probability sampling method where naturally existing clusters of a population are targeted for random selection.**

__Clustered random sampling__**- A statistical test where conditional independence of the association between two categorical variables can be assessed.**

__Cochran-Mantel-Haenszel__**- A within-subjects statistical test where three or more observations of a dichotomous categorical outcome are compared across time or within-subjects.**

__Cochran's Q__**- A document that contains information regarding the codification scheme used for variables in a database.**

__Codebook__**- In order to establish a treatment effect, the treatment must be compared to a control group. Also known as a control group, be it inactive, active, or another interventions.**

__Comparator__**- A type of validity evidence where a survey instrument correlates with a current measure or outcome, simultaneously.**

__Concurrent validity__**- A complex mathematical solution for validating theoretical or conceptual frameworks yielded from exploratory factor analysis (EFA).**

__Confirmatory Factor Analysis (CFA)__**- A type of variable that adjusts or changes the association between predictor and outcome variables.**

__Confounding variable__**- The first step of creating a survey instrument. The researcher creates a construct specification with an explicit operational definition for the construct of interest, subsequent content areas with valid citations, and a table of allocated percentages to reflect the content areas within the test.**

__Construct specification__**- A type of validity evidence that all forms of validity evidence fall under. As more types of validity evidence are generated and a measure becomes more prevalent in the literature, more construct validity evidence is assumed to exist.**

__Construct validity__**- A type of validity evidence where a survey instrument represents the knowledge or content base associated with a construct of interest.**

__Content validity__**- A type of variable where a "true zero" exists and measures of magnitude and distance can be assessed.**

__Continuous variable__**- The proportion of control participants that had the outcome of interest. CER is used to calculate other epidemiological measures.**

__Control Event Rate (CER)__**- A type of variable that is entered into a multivariate model because it can affect the association between predictor and outcome variables. The way to control for a variable is to enter it into a multivariate model along with pertinent predictor variables.**

__Control variable__**- A non-probability sampling method where observations are chosen by researchers that have access to data.**

__Convenience sampling__**- A type of validity evidence where a survey instrument positively correlates with a measure that is theoretically or conceptually similar.**

__Convergent validity__**- A bivariate measure of relationship or association between two variables of any scale of measurement.**

__Correlation__**- A type of variable that is the frequency of times that an outcome occurs. These types of continuous outcomes are naturally skewed.**

__Count variable__**- A type of quasi-experimental design where multiple treatments can be tested at the same time.**

__Counterbalanced design__**- A statistical test of survival analysis where independent groups are compared on the "time-to-event" or temporal aspects of developing a dichotomous categorical outcome, when controlling for demographic, predictor, control, and confounding variables. Cox regression is the multivariate extension of the Kaplan-Meier curve.**

__Cox regression__**- An internal consistency measure of reliability for surveys that use Likert-type response sets.**

__Cronbach's alpha____- A type of randomized design where participants are randomly allocated to either a treatment group or a control group and after receiving the intervention, are given a "washout" period. Then, the treatment group participants receive the control intervention and the control group receives the treatment intervention. This is a powerful design because participants serve as their own controls.__

**Cross-Over Randomized Design****- A type of observational research design that establishes the prevalence of an outcome in a given population. Self-report surveys are a type of cross-sectional design.**

__Cross-sectional__**– A descriptive statistic that is useful when comparing categorical variables. Better known as the 2x2 table, it serves as the basis for calculating odds ratios, relative risk, CER, EER, ARR, ARI, NNT, NNH, sensitivity, specificity, PPV, NPV, diagnostic accuracy, Cochran-Mantel-Haenszel, chi-square, and Fisher’s Exact test.**

__Cross-tabulation__### D

**- An electronic spreadsheet where researchers input, manipulate, and store research data.**

__Database__**- A type of variable that is used to describe the characteristics of a sample in regards to age, gender, race, ethnicity, socioeconomic status, or other variables of interest.**

__Demographic variable__**- The variable that is being studied or measured in a research study.**

__Dependent variable__**- A set of statistical tests that are used to describe samples from a population.**

__Descriptive statistics__**- A diagnostic testing measure that provides the overall proportion of correct diagnoses yielded from a diagnostic test.**

__Diagnostic accuracy__**- A method for establishing the ability of a diagnostic test to detect disease states or healthy people in a precise and accurate fashion.**

__Diagnostic testing__### E

**– Commonly known as the hypothesized difference between independent groups (between-subjects) or observations of an outcome across time (within-subjects). It is also the hypothesized magnitude of an association (correlation) or expected beta weight (multivariate).**

__Effect size__**- The study of disease states in populations.**

__Epidemiology__**- A type of experimental design where two treatments are tested to assess if they are "equally as good."**

__Equivalency trial__**- A quasi-experimental design where multiple "washout" periods are used to establish treatment effects.**

__Equivalent Time Sample Design__**– A framework for integrating clinical evidence into clinical practice. There are five components of the framework: Asking clinical questions, acquiring clinical evidence, appraising clinical evidence, applying clinical evidence, and assessing evidence-based practice.**

__Evidence-based medicine (EBM)____- When choosing exclusion criteria in a research study, you are focused on types of individuals at risk of being lost to follow-up, do not possess necessary demographic or clinical characteristics related to a study, or may experience adverse effects of treatment.__

**Exclusion criteria****- The proportion of treatment participants that had the outcome of interest. EER is used to calculate other epidemiological measures.**

__Experimental Event Rate (EER)__**- Step two of the eight-step validated methodology of creating survey instruments. The construct specification is given to a panel of experts and suggestions, revisions, and additions are integrated into the construct specification. The panel should be made up of experts in the area of interest.**

__Expert Review__### F

**- A type of validity evidence where a survey instrument appears, at face value, to measure what it is supposed to measure.**

__Face validity__**- A popular mnemonic for formulating research questions. The mnemonic stands for feasible, interesting, novel, ethical, and relevant.**

__FINER__**- A between-subjects statistical test where two independent groups are compared on a dichotomous categorical outcome. This statistic is used when there are less than five observations in any of the cells in the 2x2 table.**

__Fisher's Exact test__**- A type of multivariate test where continuous outcomes are assessed concurrently across multiple levels of several categorical predictor variables.**

__Fixed-effects ANOVA__**– A descriptive statistic used to describe categorical data. A frequency is the number of times that something occurs or does not occur.**

__Frequency__**- A non-parametric statistical test used for three or more observations of an ordinal outcome across time or within-subjects. It is also used when the statistical assumptions of repeated-measures ANOVA cannot be met.**

__Friedman's ANOVA__### G

**- A goal is something that a person or group of people want to achieve or obtain. There are always objectives associated with completing goals.**

__Goal__**– A statistical correction made with repeated-measures ANOVA due to violation of the assumption of sphericity.**

__Greenhouse-Geisser__### H

**- A type of regression where predictor, demographic, control, and confounding variables are entered into a model sequentially to understand their contribution of unique variance to the outcome.**

__Hierarchical regression__**- A statistical assumption that is assessed when comparing independent groups on a continuous outcome. It is tested using the Levene's test.**

__Homogeneity of variance__**- This term is the same as homogeneity of variance.**

__Homoscedasticity____- An empirical methodology that exists coincident with inferential statistics that serves as the basis for either rejecting a null hypothesis that an association exists or not rejecting a null hypothesis because an association does exist.__

**Hypothesis testing**### I

**– The number of new cases that occur moving forward in time.**

__Incidence____- Inclusion criteria specifically indentify the characteristics of participants that are going to be objects of analysis. Demogrpahic, clinical, geographical, and temporal characteristics are most often used to define inclusion criteria. With criteria, a deductive approach is used to define the specific population that is being studied.__

**Inclusion criteria****- A type of validity evidence where a survey instrument adds unique variance to the current level of measurement and understanding for a construct. Stepwise regression is used to yield evidence of incremental validity.**

__Incremental validity__**– A statistical assumption where each participant is only counted as one observation and are independent of all other participants.**

__Independence of Observations (IOO)__**- A between-subjects statistical test where two independent groups are compared on a continuous outcome, given that the assumptions of independence of observations (IOO), normality, and homogeneity of variance are met.**

__Independent samples t-test__**- A type of variable that is manipulated by researchers to better understand its association with a dependent variable.**

__Independent variable__**– A descriptive statistic used with medians to describe the variability of an ordinal distribution or a non-normal continuous distribution.**

__Interquartile range (IQR)__**- A type of quasi-experimental design where a treatment effect can be assessed across multiple measurements of the outcome.**

__Interrupted time series design__**- A part of the PICO mnemonic that describes the intervention that is being administered to study participants.**

__Intervention__**- A measure of inter-rater reliability where ratings are given at a continuous level of measurement.**

__Intraclass Correlation Coefficient (ICC)__**- A type of reliability evidence where an assumption is made that all items should be intercorrelated because they are written to cover one construct that has several component areas. If all the items correlate above a certain level (Cronbach's alpha > .75), the evidence of internal consistency reliability can be assumed. Other forms of internal consistency reliability include response sets with dichotomous responses, Likert responses, and continuous responses.**

__Internal Consistency Reliability____- A type of reliability evidence where independent raters of the same phenomenon, characteristic, or behavior are assessed on their agreement on a certain performance, event, or occurrence.__

**Inter-Rater Reliability****- A type of variable that does not possess a "true zero" but can be used with parametric statistics, given the meeting of statistical assumptions. Interval level measurement can provide measures of distance, but not magnitude.**

__Interval variable__**- The part of a survey item that acts a stimulus to elicit a reaction from participants.**

__Item stem__### J

**- A method for validating statistical findings. Each participant is removed from the sample sequentially and statistical analyses are repeated until every derivation of the sample is assessed.**

__Jack-knife__### K

**- A statistical test or survival analysis where independent groups are compared on the "time-to-event" or temporal aspects of a dichotomous categorical outcome.**

__Kaplan-Meier__**- A measure of inter-rater reliability where ratings are given at a dichotomous categorical level of measurement.**

__Kappa__**- A type of validity evidence where a survey instrument differentiates between independent groups. Between-subjects statistics are used to establish this type of validity evidence.**

__Known-groups validity__**- A between-subjects statistical test where three independent groups are compared on an ordinal outcome. The Kruskal-Wallis test is also used when the assumption of homogeneity of variance is violated for an ANOVA.**

__Kruskal-Wallis__**- An internal consistency measure of reliability for surveys that use dichotomous response sets (yes/no).**

__Kudar-Richardson 20 (KR-20)__**- A descriptive statistic that tests for normality of a continuous distribution.**

__Kurtosis__### L

**– A statistical test used to assess the assumption of homogeneity of variance or homoscedasticity for in between-subjects comparisons with continuous outcomes.**

__Levene’s test__**- A method for constructing response sets where participants rate their perceptions or reactions to an item stem along an ordered continuum.**

__Likert scale__**- A regression diagnostic of multiple regression and logistic regression where a linear relationship must exist between continuous predictor variables and the outcome variable.**

__Linearity__**– A statistical method used when continuous distributions are skewed due to outliers. The natural log (ln) of each observation is taken and inferential statistics are conducted. The mean values are no longer interpretable, but the p-value associated with the inferential statistic can be interpreted.**

__Logarithmic transformation__**- A type of regression where the outcome is a dichotomous categorical outcome variable. Adjusted odds ratios with 95% confidence intervals are the primary inferences yielded from the analysis.**

__Logistic regression__### M

**- A between-subjects statistical test where two independent groups are compared on an ordinal outcome. The Mann-Whitney U test is also used when the assumption of homogeneity of variance is violated for an independent-samples t-test. Also, the Mann-Whitney U test is used as a post hoc test for significant main effects for Kruskal-Wallis tests.**

__Mann-Whitney U__**- A framework for understanding how people develop in life and achieve their potential.**

__Maslow's Hierarchy of Needs__**– A statistical test used to assess the assumption of sphericity with repeated-measures ANOVA.**

__Mauchly’s test__**- A within-subjects statistical test where two observations of a categorical outcome are compared across time or within-subjects.**

__McNemar's__**- A descriptive statistic that is the average of a continuous distribution. All values in the distribution are added together and that value is divided by the number of observations. It is used to give context to the findings of inferential statistics and**

__Mean__*p*-values.

**- A descriptive statistic that shows what observation occurs in the middle of a continuous distribution. It is used to give context to the findings of ordinal level variables and some non-parametric statistics.**

__Median__**- A type of multivariate statistical test where continuous outcomes are compared between independent groups (between-subjects) across time or within-subjects. Essentially, researchers are testing to see if independent groups change at a different pace across time in regards to the outcome.**

__Mixed-effects ANOVA__**- A descriptive statistic that shows what observation in a continuous distribution occurs the most often.**

__Mode____- A phenomenon in regression modeling where predictor variables are highly correlated to each other and artifically inflate multivariate associations. Variance Inflation Factor (VIF) and Tolerance are used to assess multicollinearity in regression models.__

**Multicollinearity****- A type of regression where the outcome is a polychotomous categorical outcome variable. Adjusted odds ratios with 95% confidence intervals are the primary inferences yielded from the analysis.**

__Multinomial logistic regression__**- A type of regression where the outcome is a continuous variable. Statistical assumptions of normality, homoscedasticity, and linearity must be met before interpreting a multiple regression model.**

__Multiple regression__**- A type of multivariate statistical test where multiple continuous outcomes are adjusted when comparing independent groups.**

__Multivariate analysis of covariance (MANCOVA)__**- A type of multivariate statistical test where multiple continuous outcomes are compared concurrently across independent groups.**

__Multivariate analysis of variance (MANOVA)__### N

**- A method of regression used when predicting for count outcomes. The variance of the count outcome is higher than the mean of the outcome.**

__Negative binomial regression__**- The proportion of people that test negative with a diagnostic test and do not have the disease state as rated by the "gold standard."**

__Negative Predictive Value (NPV)____- An observational design embedded into a prospective cohort study where baseline measures and specimens are available for analysis.__

**Nested case-control****- A type of variable where things are "named" or "categorized" using numerical values. Computers and software programs do no understand what variables mean at a qualitative level. Therefore, nominal variables use logical numerical headings to denote group participation or possession of characteristic or trait.**

__Nominal variable____- A statistical technique where the predicted probabilities of a regression model are mapped across a scoring system to give patients and clinicians a relative understanding of the risk of developing an outcome given certain risk factors.__

**Nomograms****- A type of quasi-experimental design where randomization occurs at the point of intervention, meaning that groups are assigned to treatments.**

__Nonequivalent control group design__**- A type of experimental design where a treatment is tested to assess if it is "just as good" as the "gold standard."**

__Non-inferiority trial____- A "family" of inferential statistical tests that are used with categorical and ordinal outcomes and when the statistical assumptions of parametric statistics (normality and homogeneity of variance) are violated.__

**Non-Parametric Statistics****- A type of sampling methodology where participants are chosen by researchers for participation.**

__Non-probability sampling__**- A regression diagnostic used to assess the assumptions of normality and homoscedasticity in multiple regression. It is also known as a P-P plot. The plot is the cumulative frequency of the distribution of standardized residuals yielded from the model against the residuals associated with a normal probability graph scale.**

__Normal probability plot__**- A statistical assumption for continuous level measurements where the distribution resembles the "bell-curve," normal distribution, or Gaussian distribution. Skewness and kurtosis statistics are used to assess the assumption of normality. Skewness and kurtosis statistics below an absolute value of 2.0 are considered normal. Values above 2.0 assume a non-normal distribution.**

__Normality__**– A statistical assumption of within-subjects analyses using a continuous outcome where the differences between respective observations of the outcome are normally distributed as per skewness and kurtosis statistics.**

__Normality of difference scores__**- The number of people that have to be treated to cause a bad outcome in the future. Higher NNH values are preferable.**

__Number Needed to Harm (NNH)__**- The number of people that need to be treated to prevent a bad outcome in the future. Lower NNT values are preferable.**

__Number Needed to Treat (NNT)__### O

**- An action or behavior that upon completion, leads towards the completion of a goal. Bloom's Taxonomy is an excellent framework for writing goals and objectives.**

__Objective__**- A statistical test where an expected median value is compared to an observed median value in a population.**

__One-sample median test__**- A statistical test where an expected mean value is compared to an observed mean value in a population.**

__One-sample t-test____- A type of hypothesis that postulates a "directional" effect in either an increasing or decreasing fashion. These types of hypotheses should be used rarely. Most journals require the more rigorous two-sided hypothesis.__

**One-sided hypothesis**__- A type of outcome measured at the ordinal level. Proportional Odds Regression is used to assess this type of multivariate method.__

**Ordinal outcome****- A type of variable where a sense of order but not distance or magnitude is assessed. Likert-type scales are considered ordinal.**

__Ordinal variable__**- A type of variable that can be observed and measured in a valid fashion.**

__Outcome variable__### P

__- A type of experimental design where study participants are randomly allocated to either the treatment or control group and stay in that group throughout the entirety of the study.__

**Parallel Randomized Design**__- A "family" of inferential statistical tests that are used with continuous outcomes and when statistical assumptions like normality and homogeneity of variance are met.__

**Parametric Statistics**__- A statistical test of magnitude and direction of association between two continuous variables.__

**Pearson's r****- A statistical test of magnitude and direction of association between two categorical variables.**

__Phi-coefficient__**- A popular mnemonic for writing research questions. The mnemonic stands for population, intervention, comparator, and outcome.**

__PICO__**- The seventh step of creating a survey instrument. The pretested survey is given to participants from the population of interest (**

__Pilot study__*n*= 150 - 300). Exploratory factor analysis with principal component analysis is used to understand the underlying factor structure of the survey instrument. Cronbach's alpha is run on subscales and the aggregate instrument.

**- A statistical test of magnitude and direction of association between a categorical variable and a continuous variable.**

__Point Biserial__**- A method of regression used when predicting for count outcomes. The mean of the count outcome is higher than the variance of the outcome.**

__Poisson regression__**- The proportion of people that test positive for a diagnostic test and have the disease state as rated by the "gold standard." As prevalence for a disease state increases, the PPV will increase.**

__Positive Predictive Value (PPV)____- Precision in measurement relates to the reliability, consistency, and confidence associated with a variable or outcome.__

**Precision in measurement****- A type of validity evidence where a survey instrument can predict for future outcomes.**

__Predictive validity__**- A type of variable that is hypothesized to have an association with an outcome variable.**

__Predictor variable__**– The proportion of cases that exist in a population at a given time.**

__Prevalence____- A method for reducing data into conceptual and mathematical factors that can be assessed.__

**Principal Components Analysis****- A type of sampling methodology where all members of a given population have an equal chance of being selected for participation in an experiment.**

__Probability sampling__**– A method used in retrospective designs where cases are individually matched to similar controls in a population using a mathematical algorithm. Logistic regression and multinomial logistic regression are used to conduct propensity score matching.**

__Propensity score matching__**- A statistical assumption of proportional odds regression where the effect is stable across levels of the ordinal outcome.**

__Proportional odds__**- A type of regression used when predicting for ordinal outcomes. Adjusted odds ratios with 95% confidence intervals are the primary inference reported.**

__Proportional odds regression__**- A type of research design where the outcome of interest will occur in the future.**

__Prospective__**- A type of observational research design that establishes the relative risk, incidence, and longitudinal effects of an outcome.**

__Prospective cohort__**- A non-probability sampling method where researchers target specific groups in a population for sampling.**

__Purposive sampling__### R

**- A method used in experimental designs where study participants that have been randomly selected from the population are randomly allocated to either the treatment group or the control group.**

__Random assignment__**- A type of multivariate test where multiple within-subjects effects are tested across time.**

__Random-effects ANOVA__**- A method used in experimental designs where all members of a given population have an equal chance of being selected for participation in the study.**

__Random selection__**- An experimental research design where study participants are randomly selected from the population and randomly assigned to treatment groups. Analyses must be conducted in a "intention-to-treat" fashion.**

__Randomized Controlled Trial (RCT)____- Experimental designs where study participants are randomly assigned to treatment groups so that they are assumed to be comparable or possess equipose at baseline before an intervention.__

**Randomized research designs****- A statistical test of magnitude and direction of association between an ordinal variable and a categorical variable.**

__Rank Biserial__**- A type of variable where a "true zero" exists and measures of magnitude and distance can be assessed.**

__Ratio variable__**- The consistency, stability, and precision of variables and measurement.**

__Reliability__**- A within-subjects statistical test where two observations of a continuous outcome are compared across time or within-subjects.**

__Repeated-measures t-test__**- A within-subjects statistical test where three or more observations of a continuous outcome are compared across time or within-subjects.**

__Repeated-measures ANOVA__**- A type of regression diagnostic that provides a measure of model fit. Residuals are the difference between the observed value and the predicted value yield from a multivariate model.**

__Residual analysis__**- A part of a survey item where participants give ratings based on their reactions or perceptions to the item.**

__Response set__**- A type of research design where the outcome of interest occurred in the past.**

__Retrospective__**- A type of observational research design that can yield measures of relative risk and longitudinal effects. A cohort is selected from a given population and the risk of developing an outcome based on exposure/non-exposure to a predictor variable.**

__Retrospective cohort__### S

**– The absolute number, or**

__Sample size__*n*, of individuals selected for participation in the study. Samples are derived from populations and inferential statistics allows us to make generalizations back to the population.

__- A post hoc test that is considered one of the strongest against deterring increased experimentwise error rates when adjusting for multiple comparisons.__

**Scheffe's test****- A survey methodology where respondents give their perceptions or beliefs related to a construct.**

__Self-report__**- The ability of a diagnostic test to detect disease states. When a diagnostic test has a sensitivity greater than 95%, the test can "rule out" disease states.**

__Sensitivity__**- A probability sampling technique where every member of the population has an equal chance of being chosen for participation in a study.**

__Simple random sampling__**- A multivariate model where all relevant predictor, demographic, control, and confounding variables are entered concurrently into the model to assess multivariate associations.**

__Simultaneous regression__**- A descriptive statistic that tests for normality of a continuous distribution.**

__Skewness__**- A statistical test of magnitude and direction of association between two ordinal variables.**

__Spearman's rho__**- The ability of a diagnostic test to identify healthy people. When a diagnostic test has a specificity greater than 95%, the test can "rule in" disease states.**

__Specificity__**– A statistical assumption of repeated-measures ANOVA that is often violated. Greenhouse-Geisser corrections are used to correct for this prevalent violation.**

__Sphericity__**- A method for validating statistical findings. The sample is randomly divided in half and psychometrics are run on the respective halves (the derivation sample and the confirmatory sample). The groups can be randomly split into percentages of 50/50, 60/40, or 70/30. If the findings on each sample are similar, then evidence of split-group validity is assumed.**

__Split-group__**- An internal consistency measure of reliability where two independent halves of a survey instrument are significantly correlated.**

__Split-half reliability__**– A descriptive statistic that gives a relative context for understanding how far away individual observations are away from the mean in a continuous distribution.**

__Standard deviation__**- A multivariate model where an algorithm chooses the best combination of predictor, demographic, control, and confounding variables to predict for an outcome.**

__Stepwise regression__**- A probability sampling technique where defined subgroups or strata of a population can be targeted for representation in a random sample.**

__Stratified random sampling__**- A cross-sectional methodology where participants are asked to respond to questions regarding a construct of interest.**

__Survey__**- The methods by which participants are given a survey instrument. There are five primary modes: One-on-one, group, telephone, postal mail, and electronic.**

__Survey modes of administration__**- The structure of survey as it is presented to participants. The six parts of a survey are the title, introduction, instructions, survey items, demographics, and closing statement.**

__Survey parts__**- The type of survey used is dependent upon the research question being asked. There are six types of surveys: Test, rating scale, performance, checklist, psychological instrument, and inventory.**

__Survey types__**- A set of statistical tests that compare independent groups on the temporal aspects of developing an outcome. The primary types of survival analysis are Kaplan-Meier and Cox regression.**

__Survival analysis__**- The highest level of applied clinical evidence. The results of multiple randomized controlled trials are aggregated according to rigorous inclusion and exclusion criteria and interpreted as a pooled effect. Meta-analysis is the statistical technique used with systematic reviews. The results of a meta-analysis are reported in a forest plot. Heterogeneity of the effect is reported with the I2 statistic. A funnel plot is used to assess publication bias in the respective randomized controlled trials.**

__Systematic review__### T

**- A type of reliability evidence that assesses the consistency and stability of a survey instrument outcome across time.**

__Test-retest reliability____- A method for assessing levels of multicollinearity in regression models. Lower values of tolerance denote increased multicollinearity.__

**Tolerance**__- A post hoc test that is popular in the social sciences when adjusting for multiple comparisons and increased experimentwise error rates. HSD stands for Honestly Significant Difference.__

**Tukey's HSD test**__- A type of hypothesis that does not stipulate if an effect will be either positive or increasing nor negative or decreasing. These types of hypotheses are required in most scientific journals because they are more rigorous.__

**Two-sided hypothesis**__- An error that occurs in hypothesis testing when the null hypothesis is rejected when it should not be rejected (false positive). The alpha value or significance value chosen in a study is the chance that a researcher is willing to take for committing a Type I error and is most often times set at .05.__

**Type I error**__- An error that occurs in hypothesis testing when the null hypothesis is not rejected when is should be rejected (false negative. The beta value represents the chance that a researchers is willing to take for committing a Type II error. Type II errors typically occur when there are not observations of an outcome.__

**Type II error**__- An error that occurs in hypothesis testing when the wrong question is answered by a statistical test. Type III errors often occur due to miscommunication between statisticians and researchers and also when mining data and not adjusting for multiple comparisons.__

**Type III error**### U

__- A method of randomization where participants are randomly allocated to treatment groups in an uneven fashion. This method of randomization is used when the intervention itself is of particular interest. Ratios of allocation of 2:1, 3:1, and 4:1 to treatment groups are acceptable. Unequal allocation randomization is used to study more serious disease states that require stronger treatments. The side effects of promising treatments can also be studied using unequal allocation.__

**Unequal allocation randomization**### V

**- The utility, interpretability, and accuracy of outcomes and measurement.**

__Validity__**- The use of jack-knife, split-group, and bootstrap methods to improve the precision and accuracy of statistical findings.**

__Validation of statistical findings__**- The eighth and final step of creating a survey instrument. The piloted survey is given to a new sample (**

__Validation study__*n =*300 - 1,000) participants from the population of interested. Validity coefficients and confirmatory factor analysis is run.

**– A descriptive statistic explaining the degree of dispersal of observations in a distribution of continuous values.**

__Variance____- A method for assessing levels of multicollinearity in a regression model. Higher values of VIF denote increased multicollinearity in a model.__

**Variance Inflation Factor (VIF)**### W

**- A within-subjects statistical test for two observations of an ordinal outcome across time or within-subjects.**

__Wilcoxon test__**- A statistical method for comparing observations across time or within the same group.**

__Within-subjects__### Y

__- A measure of the overall ability of a diagnostic test. The formula is (Senstivity + Specificity) - 100. Any value below 50 denotes a poor diagnostic test.__

**Youden index**Click on a button below to continue.