Psychological Statistics: Beginner-Friendly Reviewer
Foundations of Variables and Levels of Measurement
Definition and Scope of Statistics in Psychology:
- Statistics represents the systematic procedure of collecting, organizing, analyzing, and interpreting information or data.
- Within psychological science, statistics provides empirical and numerical tools to evaluate human behavior, emotional experiences, cognitive processes, and personal phenomena.
- Applied Example: A researcher evaluating student anxiety gathers self-report scores from a cohort, organizes the distribution, computes the mean anxiety level, and draws conclusions regarding the sample's emotional state.
Core Branches of Statistical Analysis:
- Descriptive Statistics:
- Focuses strictly on summarizing, organizing, and characterizing the specific data gathered.
- Common tools include measures of central tendency (, , ), measures of dispersion (, ), and frequency tallies.
- Applied Example: Calculating an average anxiety score of from an evaluated group of students serves to describe that specific cohort without generalizing beyond it.
- Conceptual Summary: Descriptive statistics describes the sample data at hand.
- Inferential Statistics:
- Focuses on utilizing sample statistics derived from a smaller cohort to make probabilistic inferences, estimates, or conclusions about a broader target population.
- Applied Example: Taking the anxiety findings from surveyed university students to evaluate whether those trends generalize to the broader university student population.
- Conceptual Summary: Inferential statistics uses observed sample data to deduce population properties.
Definition and Classification of Variables:
- A variable is any property, characteristic, or attribute that can assume different values across individuals, settings, or time.
- Representative Examples: Age, gender, anxiety level, depression score, academic performance, hours spent on TikTok, and sleep duration.
- Operational Rationale: It is termed a variable because individual observations vary (e.g., Student A sleeps for hours, Student B sleeps for hours, and Student C sleeps for hours).
Classification of Variables by Research Role:
- Independent Variable ():
- The variable manipulated by the experimenter or used as an explanatory predictor to evaluate its influence on an outcome.
- Applied Example: In a study examining whether sleep duration influences academic performance, sleep duration acts as the independent variable ().
- Dependent Variable ():
- The outcome or response variable measured to determine the effect of the predictor or experimental manipulation.
- Applied Example: In the sleep duration and academic performance study, academic performance represents the dependent variable ().
- Methodological Caveat: In non-experimental and correlational designs, variables are measured without active manipulation; identifying an explanatory variable as a predictor does not inherently imply causal directionality.
Classification of Variables by Nature:
- Qualitative (Categorical) Variables:
- Variables that classify individuals into discrete groups, labels, or nominal categories rather than quantifying magnitude.
- Examples: Gender category, college academic major, religious affiliation, marital status, and classification of psychotherapy modality.
- Quantitative (Numerical) Variables:
- Variables that represent measured amounts, counts, or continuous numerical quantities.
- Examples: Age, height, standardized anxiety scores, sibling count, and study hours.
- Discrete Quantitative Variables: Numerical values limited to specific countable quantities, typically whole numbers, with no intermediate points (e.g., number of siblings, class enrollment, absence counts; an individual cannot possess siblings).
- Continuous Quantitative Variables: Numerical measurements that can fall along an infinite continuum and assume fractional or decimal values within a given range (e.g., height, weight, reaction time, and sleep duration such as hours).
The Four Hierarchical Levels of Measurement:
- Nominal Level:
- Categorical classification based solely on names, labels, or arbitrary identifiers.
- Lacks inherent rank, numerical value, or quantitative distance between categories.
- Examples: Gender, ABO blood types, religious categories, and college degree programs (e.g., Psychology, Biology, Political Science).
- Ordinal Level:
- Involves categories that maintain a clear, logical rank order or relative hierarchy.
- The quantitative intervals between adjacent ranks are unknown or unequal.
- Examples: Ordered ratings (Low, Moderate, High), Likert satisfaction responses (Dissatisfied, Neutral, Satisfied), and competitive rankings (, , ).
- Interval Level:
- Ordered numerical measurements with equal, standardized distances between scale units.
- Lacks an absolute or true zero point; a value of zero does not signify the absolute absence of the underlying construct.
- Examples: Temperature in Celsius () and Fahrenheit (). The difference between and equals the difference between and (), yet does not denote the total absence of thermodynamic heat.
- Ratio Level:
- Features equal, standardized intervals between numerical units and incorporates a true, non-arbitrary absolute zero.
- A score of zero signifies the complete absence of the property being measured, allowing for meaningful multiplicative and ratio comparisons.
- Examples: Chronological age, physical height, body mass, reaction latency, and cumulative study time ( hours indicates complete absence of study time).
- Measurement Hierarchy Comparison Matrix:
- Nominal: Has Order = No; Equal Intervals = No; True Zero = No (Memory key: Name only).
- Ordinal: Has Order = Yes; Equal Intervals = No; True Zero = No (Memory key: Order).
- Interval: Has Order = Yes; Equal Intervals = Yes; True Zero = No (Memory key: Equal intervals).
- Ratio: Has Order = Yes; Equal Intervals = Yes; True Zero = Yes (Memory key: Real zero).
Operational Definitions in Psychological Research:
- An operational definition explicitly specifies the exact procedural operations, empirical indicators, or instruments used to quantify an abstract construct.
- Because psychological attributes (e.g., anxiety, intelligence) cannot be directly inspected physically, operationalization bridges abstract conceptual definitions and concrete empirical data.
- Applied Example: Operationalizing anxiety as the composite score obtained on a validated, standardized self-report anxiety inventory.
Frequency Distributions and Graphical Data Representations
Definition and Organization of Frequency Distributions:
- A frequency distribution tabulates the count of occurrences for each distinct value or score range observed within a dataset.
- Tabular Example: A cohort of students reports the following raw anxiety scores: .
- Score : Frequency =
- Score : Frequency =
- Score : Frequency =
- Score : Frequency =
- Score : Frequency =
- Score : Frequency =
- Total Observations () =
- The modal score is , exhibiting the maximum frequency ().
Mathematical Components of Frequency Tables:
- Absolute Frequency ():
- The direct count of times an exact numerical score or category appears.
- Relative Frequency ():
- The proportion of the overall dataset represented by an individual score value.
- Mathematical Formulation:
- Where represents the specific frequency and represents total observations.
- Percentage Frequency:
- The relative frequency expressed as a percentage of the total sample.
- Mathematical Formulation:
- Computational Example: If out of participants endorse "Agree":
- Cumulative Frequency ():
- The progressive running sum of frequencies up to and including the current score or class interval.
- Progressive Calculation Example:
- Score : ,
- Score : ,
- Score : ,
- Score : ,
- Score : ,
- Score : ,
- Class Intervals:
- Standardized groupings of continuous or discrete numerical ranges designed to compress wide distributions (e.g., , , ).
- Class Midpoint:
- The precise center value of an established class interval.
- Mathematical Formulation:
- Computational Example: For the interval :
Graphical Data Displays:
- Bar Graph:
- Depicts categorical or qualitative discrete data using separated bars to reflect qualitative boundaries.
- Distinctive feature: Distinct spaces are maintained between vertical bars.
- Histogram:
- Depicts continuous numerical data grouped into equal class intervals.
- Distinctive feature: Adjacent bars touch one another directly to signify underlying data continuity.
- Frequency Polygon:
- Employs data points plotted above class midpoints and connected by line segments to outline distribution contour and shape.
- Ogive:
- A cumulative frequency line graph showing the accumulation of frequencies across score thresholds or class upper limits.
- Pie Chart:
- A circular graphic segmented into sectors proportional to the relative percentages or proportions of a total categorical variable.
- Summary of Visual Displays:
- Bar Graph: Depicts distinct categories or discrete classifications.
- Histogram: Visualizes grouped continuous or interval data.
- Frequency Polygon: Delineates frequency trajectories and overall distributional form.
- Ogive: Illustrates running cumulative frequencies.
- Pie Chart: Demonstrates proportional parts of an aggregate whole.
Descriptive Statistics: Percentiles, Central Tendency, and Dispersion
Percentile Ranks and Quartile Divisions:
- Percentiles:
- Quantify the relative standing or position of a single score by stating the percentage of total observations that fall at or below it.
- Applied Interpretation: A score at the percentile indicates that an individual's performance meets or exceeds approximately of the comparison cohort.
- Conceptual Distinction: Percentile rank reflects positional standing within a distribution, not the raw percentage of test items answered correctly.
- Standard Percentile Benchmarks: ( at or below), ( at or below), ( at or below), ( at or below).
- Quartiles:
- Partition an ordered array of scores into four equal segments, demarcated by three cutting points:
- First Quartile (): Corresponds to ; demarcates the lower of data.
- Second Quartile (): Corresponds to ; identical to the distribution .
- Third Quartile (): Corresponds to ; demarcates the boundary below which of observations lie.
- Interquartile Range ():
- Measures statistical dispersion across the central of an ordered distribution.
- Mathematical Formulation:
- Computational Example: Given and :
- Substantive Interpretation: The middle of observations spans a range of measurement units.
Measures of Central Tendency:
- Arithmetic Mean:
- Represents the balance point or mathematical average of all values.
- Mathematical Formulation:
- Step-by-Step Computational Example: Given raw scores :
- Step 1 (Summation):
- Step 2 (Count):
- Step 3 (Division):
- Analytical Strengths: Utilizes all numerical values in the dataset; mathematically tractable for advanced inferential procedures.
- Analytical Weaknesses: Shows high sensitivity to extreme values (outliers) and can become unrepresentative in asymmetrical, skewed distributions.
- Median:
- The physical midpoint separating an ordered distribution into equal halves ( above, below).
- Odd Number Calculation: For ordered series , the middle score is .
- Even Number Calculation: For ordered series , the two center scores ( and ) are averaged:
- Analytical Strengths: Robust against extreme outliers; represents skewed data effectively; valid for ordinal-level metrics.
- Analytical Weaknesses: Ignores the precise numerical magnitude of individual values flanking the midpoint.
- Mode:
- The specific value, score, or category exhibiting the highest frequency in a dataset.
- Numerical Example: In array , the modal value is ().
- Modal Classifications: Unimodal (one peak), Bimodal (two distinct peaks), Multimodal (more than two peaks), No Mode (every value occurs with identical frequency).
- Comparative Selection Framework:
- Mean: Optimal for symmetrical, approximately normal quantitative datasets devoid of severe outliers.
- Median: Optimal for skewed numerical distributions, datasets containing outliers, or purely ordinal data.
- Mode: Optimal for nominal categorical data or identifying the single most common category/score.
- Data Scale Alignment:
- Nominal: Mode only.
- Ordinal: Median or Mode.
- Interval/Ratio: Mean, Median, or Mode (conditioned by distributional symmetry and outlier presence).
Measures of Variability (Dispersion):
- Conceptual Importance: Variability quantifies the degree of spread, diversity, or dispersion of scores around the central value. Two groups can possess identical means while exhibiting entirely distinct dispersion profiles (e.g., Class A with scores vs. Class B with scores ; both have , yet Class B demonstrates substantially wider spread).
- Range:
- The gross distance between the absolute extremes of a distribution.
- Mathematical Formulation:
- Computational Example: For values :
- Limitation: Sensitive exclusively to the two extreme endpoints, ignoring the distribution of intermediate scores.
- Variance:
- Quantifies average squared distance of individual scores from their mean.
- Population Variance Formulation:
- Sample Variance Formulation:
- Symbol Definitions: = raw score, = population mean, = sample mean, = population size, = sample size, = summation notation.
- Rationale for Bessel's Correction (): Utilizing sample deviations to estimate population dispersion systematically underestimates true population variance. Dividing by degrees of freedom () corrects this negative bias, providing an asymptotically unbiased estimator of .
- Standard Deviation ():
- The positive square root of variance, restoring the metric of dispersion to original measurement units.
- Mathematical Formulation:
- Interpretive Heuristics:
- Small : Scores cluster tightly around the arithmetic mean.
- Large : Scores spread broadly across the scale.
- : Every individual observation in the dataset holds the identical numerical score.
- Applied Comparison: Group A () versus Group B () demonstrates that Group B exhibits substantially greater variability.
- Dimensional Distinction: Variance is quantified in squared units; standard deviation is quantified in original baseline units.
- Coefficient of Variation ():
- A standardized measure of relative dispersion describing the ratio of standard deviation to mean expressed as a percentage.
- Mathematical Formulation:
- Applied Utility: Facilitates direct comparisons of relative variability across scales with disparate measurement units, provided valid ratio-level zero points exist.
The Normal Distribution, Empirical Rule, and Distribution Shape
Theoretical Characteristics of the Normal Distribution:
- Manifests as a perfectly symmetrical, unimodal bell-shaped mathematical curve.
- The central tendencies coincide exactly at the centerline: .
- Score density clusters symmetrically near the center, diminishing smoothly toward both extreme tails.
- The total area beneath the density curve equals precisely ( probabilistic space).
- The tails approach the horizontal abscissa asymptotically, extending infinitely in both directions without ever contacting the axis.
The Empirical Rule ( Rule):
- Delineates predictable percentage bands under a normal distribution:
- Approximately of observations lie within standard deviation of the mean: .
- Approximately of observations lie within standard deviations of the mean: .
- Approximately of observations lie within standard deviations of the mean: .
- Applied Computation (Test Metric with , ):
- Range: to ; encompasses approximately of the scores.
- Range: to ; encompasses approximately of the scores.
- Range: to ; encompasses approximately of the scores.
Standardized Scores (-Scores):
- Quantifies the signed distance of a raw score from its mean in standard deviation units.
- Mathematical Formulation:
- Computational Example: Given raw score , , :
- Step 1 (Deviation):
- Step 2 (Standardization):
- Substantive Meaning: The raw performance of falls precisely standard deviations above the population mean.
- Interpretive Scale:
- : Two standard deviations above the mean.
- : One standard deviation above the mean.
- : Coincides directly with the mean.
- : One standard deviation below the mean.
- : Two standard deviations below the mean.
Distributional Skewness:
- Quantifies the degree of horizontal asymmetry present in a frequency distribution.
- Positive Skew (Right-Skewed):
- The elongated tail stretches toward the positive (right) side of the scale; bulk data concentrate on the left.
- Typical Central Tendency Hierarchy: .
- Applied Scenario: Household income distributions, where a small minority of extreme earners pulls the mean to the right.
- Negative Skew (Left-Skewed):
- The elongated tail stretches toward the negative (left) side of the scale; bulk data concentrate on the right.
- Typical Central Tendency Hierarchy: .
- Applied Scenario: An exceptionally easy classroom exam, where most students achieve high scores and a few low scores pull the tail left.
- Symmetrical Distribution:
- Balanced mirror-image structure across the centerline: .
Distributional Kurtosis:
- Reflects the relative heaviness or weight of the distribution tails relative to a Gaussian baseline.
- Mesokurtic: Tail heaviness and peakedness align with a normal distribution (Jamovi excess kurtosis ).
- Leptokurtic: Displays heavier, fatter tails and a heightened rate of extreme outliers (Jamovi excess kurtosis ).
- Platykurtic: Displays lighter, thinner tails with fewer extreme scores (Jamovi excess kurtosis ).
Foundations of Inferential Statistics and Hypothesis Testing
Core Inferential Terminology:
- Population: The complete aggregate of all potential units, subjects, or observations under conceptual review (e.g., all college students in a university).
- Sample: A finite subset selected systematically from the population to participate in empirical measurement (e.g., surveyed university students).
- Parameter: A fixed numerical index summarizing a population attribute (e.g., true population mean , population standard deviation ).
- Statistic: A variable numerical index computed from sample data (e.g., sample mean , sample standard deviation ).
- Sampling: The procedural methodology implemented to extract representative participants from a defined population.
- Sampling Error: The naturally occurring mathematical discrepancy between a sample statistic and its true underlying population parameter arising from random sampling variability.
Formal Hypothesis Specification:
- Null Hypothesis ():
- Asserts that no real difference, effect, or relationship exists in the population; observed patterns are attributed solely to random sampling fluctuation.
- Example: There is no significant relationship between fear of missing out (FOMO) and doomscrolling ().
- Alternative Hypothesis ( or ):
- The substantive experimental hypothesis proposing the genuine presence of an effect, systematic group difference, or relationship in the population.
- Example: There is a significant relationship between FOMO and doomscrolling ().
The -Value and Alpha Thresholds:
- Statistical Definition of the -Value:
- The conditional probability of observing a sample statistic as extreme as, or more extreme than, the observed data, assuming that the null hypothesis () is strictly true.
- Alpha Level ():
- The pre-specified significance criterion representing the maximum acceptable probability of committing a Type I error (standard convention: , or ).
- Decision Rules for Hypothesis Testing:
- If : Reject . The observed pattern is statistically significant, providing adequate empirical evidence to dispute the null premise.
- If : Fail to reject . The finding is statistically non-significant, yielding insufficient evidence against the null premise.
- Evaluative Table of Decisions:
- : Reject (Statistically significant).
- : Reject (Statistically significant).
- : Reject (Statistically significant).
- : Reject (Statistically significant under criterion).
- : Fail to reject (Not statistically significant).
- : Fail to reject (Not statistically significant).
- Crucial Methodological Interpretations:
- Rejecting does not prove the alternative hypothesis is absolute fact.
- Failing to reject does not prove the null hypothesis is true.
- A non-significant result does not prove the absolute absence of an effect; it indicates a failure to detect one.
- A -value does not measure the probability that itself is true.
- Statistical significance is distinct from substantive, clinical, or real-world importance.
Confidence Intervals ():
- A calculated interval from sample data that defines a plausible range of values for an unknown population parameter at a specified confidence level.
- Conceptual Meaning of a Confidence Level: If identical sampling procedures were iterated infinitely, approximately of calculated intervals would span the true population parameter.
- Applied Example: An empirical sample mean anxiety score of with a establishes that values between and constitute plausible bounds for the population mean.
- Relation to Hypothesis Testing: For two-tailed testing at , a confidence interval excludes the null hypothesis reference parameter if and only if the test is statistically significant ().
Statistical Power:
- The operational probability of correctly rejecting a false null hypothesis (detecting a genuine population effect when one exists).
- Mathematical Formulation:
- Where denotes the conditional probability of committing a Type II error.
- Determinants of Statistical Power: Sample size (), magnitude of the population effect size, choice of alpha (), sample variability, and test directionality (one-tailed versus two-tailed).
Correlational Statistics and Measures of Association
Conceptual Foundations of Correlation:
- Evaluates whether, and in what manner, two continuous or ranked variables covary systematically without experimental manipulation.
- Establishes three key properties: presence of association, magnitude/strength of association, and directional orientation.
- Applied Example: Determining whether higher student scores on FOMO scales systematically correspond to increased or decreased doomscrolling indices.
- The Causation Fallacy: Correlation does not establish causation. If screen time on TikTok correlates with FOMO scores, three distinct scenarios remain equally plausible:
- TikTok screen time actively elevates FOMO.
- Heightened FOMO drives individuals to use TikTok more frequently.
- An unmeasured confounding variable (e.g., social comparison tendency) drives both behaviors.
The Correlation Coefficient ():
- Bounded strictly within the closed numerical interval: .
- Positive Correlation ():
- Variables exhibit direct linear movement; higher values on one variable track higher values on the second.
- Example: Increased study hours correlate with higher exam marks (, indicating strong positive association).
- Negative Correlation ():
- Variables exhibit inverse linear movement; as one variable increases, the other declines.
- Example: Elevated psychological stress correlates with reduced sleep hours (, indicating strong negative association).
- Zero Correlation ():
- Denotes the absence of an identifiable linear relationship.
- Critical Caveat: An near zero does not preclude the existence of a strong non-linear or curvilinear relationship.
Quantitative Strength Benchmarks:
- Strength is determined solely by the absolute magnitude of the coefficient (), whereas direction is dictated by the algebraic sign:
- : Very weak association.
- : Weak association.
- : Moderate association.
- : Strong association.
- : Very strong association.
- Applied Examples:
- : Very strong positive correlation.
- : Very strong negative correlation.
- : Weak positive correlation.
- : Weak negative correlation.
Specific Bivariate Correlational Techniques:
- Pearson's Product-Moment Correlation ():
- Parametric metric indexing the degree of linear relationship between two quantitative, interval- or ratio-level variables.
- Underlying Assumptions: Continuous data, bivariate normality, linear relationship, homoscedasticity, independent observations, and absence of influential outliers.
- Spearman's Rank-Order Correlation (Spearman's or ):
- Non-parametric metric quantifying monotonic associations using ranked data.
- Evaluates whether variables increase or decrease together regardless of whether the trend is linear.
- Primary Utilities: Applied when variables are measured at the ordinal level, when distributions deviate from normality, or when monotonic outliers distort Pearson's metric.
- Kendall's Tau ():
- Non-parametric metric measuring rank association based on concordant and discordant pairs.
- Primary Utilities: Preferred for small sample sizes, distributions characterized by high rates of tied ranks, and non-parametric rank comparisons.
- Comparative Methodological Matrix:
- Pearson's : Parametric; evaluates linear bivariate trends in quantitative data.
- Spearman's : Non-parametric; evaluates monotonic trends across ranked variables.
- Kendall's : Non-parametric; evaluates concordance and discordance among tied or small-sample ranks.
Directional vs. Non-Directional Hypotheses and Tail Tests
One-Tailed (Directional) Testing Framework:
- Specifies the exact directional path (increase or decrease) of an outcome prior to data collection.
- Places the entire alpha rejection region () into a single designated tail of the test distribution.
- Right-Tailed Structure:
- Hypothesizes that the focal parameter is greater or higher than the baseline.
- Hypotheses: versus .
- Example: Predicting that an experimental study intervention will produce higher test scores than a control condition.
- Left-Tailed Structure:
- Hypothesizes that the focal parameter is lower or smaller than the baseline.
- Hypotheses: versus .
- Example: Predicting that mindfulness training will produce lower anxiety scores.
Two-Tailed (Non-Directional) Testing Framework:
- Evaluates whether a significant difference or relationship exists without specifying the direction of the outcome.
- The significance region () is partitioned equally across both tails of the reference distribution ( in the left tail, in the right tail).
- Hypotheses: versus .
- Example: Investigating whether mindfulness training changes anxiety scores without specifying whether they will increase or decrease.
Directional vs. Non-Directional Linguistic Markers in Hypotheses:
- Directional Markers (Requires One-Tailed Test if specified a priori):
- Higher than
- Lower than
- Greater than
- Less than
- Positively related
- Negatively related
- Non-Directional Markers (Requires Two-Tailed Test):
- Different from
- Associated with (unless explicit directional sign is provided)
- Related to (unless explicit directional sign is provided)
Decision Criteria for Selecting Test Tails:
- Use a One-Tailed Test when: A specific directional outcome is justified by empirical theory and decided prior to examining data, and an effect in the unpredicted direction would be functionally equivalent to no effect.
- Use a Two-Tailed Test when: The direction of the outcome cannot be predicted with certainty, exploration of effects in both directions is scientifically meaningful, or hypotheses are non-directional.
- Caution: One-tailed tests should never be adopted after inspecting the data simply because a directional trend appears favorable.
Hypothesis Testing Decisions, Errors, and Statistical Power
The Decision Framework for the Null Hypothesis:
- Rejecting the Null Hypothesis ():
- Indicates that the sample data deviate sufficiently from the null model to warrant its rejection.
- Example Script for Significant Outcome: "The results revealed a statistically significant relationship between FOMO and doomscrolling (). Therefore, the null hypothesis is rejected."
- Failing to Reject the Null Hypothesis ():
- Indicates that the sample data do not diverge sufficiently from the null premise to rule out random fluctuation.
- Avoid Saying "Accept the Null": A non-significant result never proves the null hypothesis is true; it merely indicates insufficient statistical evidence to reject it (e.g., due to inadequate sample size, high variability, or small effect size).
- Example Script for Non-Significant Outcome: "The results revealed no statistically significant relationship between FOMO and doomscrolling (). Therefore, the null hypothesis is not rejected." (Or: "The study did not find sufficient evidence of a statistically significant relationship between the variables.")
- Tabular Mapping of -Values to Decisions:
- : Reject (Significant)
- : Reject (Significant)
- : Reject (Significant)
- : Reject (Significant)
- : Fail to reject (Not significant)
- : Fail to reject (Not significant)
- : Fail to reject (Not significant)
Type I and Type II Errors:
- Type I Error ():
- Occurs when a researcher rejects a null hypothesis that is true in the population.
- Represents a "False Positive" conclusion (inferring an effect or difference that does not exist).
- Probability of Occurrence: Equal to ; under standard , the long-run risk of a Type I error across true null models is .
- Applied Example: Concluding an intervention significantly lowers anxiety when it produces no actual population change.
- Type II Error ():
- Occurs when a researcher fails to reject a null hypothesis that is false in the population.
- Represents a "False Negative" conclusion (missing an effect or difference that genuinely exists).
- Probability of Occurrence: Designated as .
- Applied Example: Concluding an intervention has no significant effect when it actually produces meaningful improvement.
- Formal Decision Matrix:
- Null Hypothesis is True & Reject : Type I Error (Probability = ).
- Null Hypothesis is True & Fail to Reject : Correct Decision (Probability = ).
- Null Hypothesis is False & Reject : Correct Decision / Statistical Power (Probability = ).
- Null Hypothesis is False & Fail to Reject : Type II Error (Probability = ).
Interrelations Between Errors and Power:
- Lowering alpha (e.g., from to ) reduces the risk of a Type I error, but increases the probability of a Type II error () if all other variables remain constant.
- Increasing sample size () reduces standard error, improves statistical power (), and decreases Type II error risk without inflating Type I error.
Testing Statistical Assumptions in Jamovi
Overview and Software Environment:
- Jamovi is an open-source statistical platform designed for data analysis, hypothesis testing, assumption evaluation, and group comparisons.
- Official portal:
https://www.jamovi.org/ - Purpose of Assumptions: Methodological conditions that must be met for a given inferential test to produce statistically valid parameter estimates and control error rates properly.
Assumption of Normality:
- Concept: Scores on the dependent variable (or model residuals) follow an approximately normal distribution within each evaluation group.
- Jamovi Execution Path:
- Analyses Exploration Descriptives.
- Move the continuous variable into the Variables box.
- If comparing discrete groups, insert the grouping variable into Split by.
- Under Statistics, select Shapiro-Wilk.
- Under Plots, activate Q-Q Plot, Histogram, and Box Plot.
- Statistical Interpretation of the Shapiro-Wilk Test:
- : No statistically significant departure from normality; the assumption is tenable.
- : Statistically significant departure from normality; the distribution deviates from Gaussian structure.
- Benchmark Scenarios: indicates tenable normality; indicates violation of normality.
- Diagnostic Caveat: Shapiro-Wilk is sensitive to sample size; large samples can yield a significant -value for minor deviations. Therefore, formal tests should be interpreted alongside Q-Q plots, histograms, and outlier screenings.
Assumption of Homogeneity of Variance:
- Concept: The comparison groups exhibit approximately equal variances on the measured outcome variable ().
- Jamovi Execution Path:
- Analyses T-Tests Independent Samples T-Test.
- Transfer outcome to Dependent Variables; transfer categorical variable to Grouping Variable.
- Under Assumption Checks, activate Homogeneity test (Levene's).
- Statistical Interpretation of Levene's Test:
- : Variances do not differ significantly; equal variance assumption is met.
- : Significant variance inequality; equal variance assumption is violated.
- Corrective Action: If Levene's , use Welch's t-test instead of Student's t-test. A non-parametric test is not automatically required when variances are unequal.
Assumption of Linearity:
- Concept: The bivariate relationship between two continuous variables follows a straight-line pattern across the score range.
- Jamovi Execution Path:
- Analyses Exploration Descriptives Scatterplot options (or dedicated scatterplot modules).
- Interpretation: Data points should cluster along a linear path. If the points curve systematically, Pearson's is inappropriate, as it indexes only linear covariation.
Identification and Management of Outliers:
- Concept: Extreme observations that deviate markedly from the overall distribution.
- Outliers can distort the Mean, inflate Standard Deviation, and bias correlation and regression estimates.
- Jamovi Boxplot Rule for Outliers:
- Lower Inner Fence:
- Upper Inner Fence:
- Observations falling outside these fences are flagged as potential outliers.
- Analytic Principle: Outliers should be investigated rather than discarded automatically, as they may represent genuine observations.
Assumption of Independence of Observations:
- Concept: Each measured observation is independent of every other observation, meaning one participant's response cannot influence another's.
- Verification: Independence is guaranteed through study design and random sampling, not through a statistical test in Jamovi.
- If the same participants are measured across multiple time points, observations are dependent, requiring paired-samples or repeated-measures designs.
Parametric vs. Non-Parametric Tests and Test Selection Guide
Core Principles of the Parametric/Non-Parametric Split:
- Parametric Tests:
- Require continuous (interval or ratio) data, independent observations, and satisfaction of distributional assumptions (e.g., approximate normality, homogeneity of variance).
- Focus on estimating population parameters and comparing means.
- Provide greater statistical power than non-parametric alternatives when their assumptions are met.
- Non-Parametric Tests:
- Make fewer distributional assumptions and do not require normal distributions.
- Commonly analyze ordinal data or transform raw scores into ranks.
- Provide robust alternatives when parametric assumptions are violated or sample sizes are small.
- Caveat: Non-parametric tests are not entirely free of assumptions; requirements such as independence of observations and specific distributional conditions still apply.
Corresponding Parametric and Non-Parametric Tests:
- Comparing Two Independent Groups:
- Parametric: Independent-Samples t-test.
- Non-Parametric Counterpart: Mann-Whitney U test.
- Comparing Two Related/Matched Measurements:
- Parametric: Paired-Samples t-test.
- Non-Parametric Counterpart: Wilcoxon signed-rank test.
- Comparing Three or More Independent Groups:
- Parametric: One-Way Analysis of Variance (One-Way ANOVA).
- Non-Parametric Counterpart: Kruskal-Wallis test.
- Comparing Three or More Related/Repeated Measurements:
- Parametric: Repeated-Measures ANOVA.
- Non-Parametric Counterpart: Friedman test.
- Evaluating Bivariate Association Between Quantitative Variables:
- Parametric: Pearson's Product-Moment Correlation ().
- Non-Parametric Counterpart: Spearman's Rank-Order Correlation () or Kendall's Tau ().
Execution and Interpretation of Specific Statistical Tests in Jamovi:
- Independent-Samples t-Test:
- Purpose: Compares the means of two distinct, independent groups.
- Navigation: Analyses T-Tests Independent Samples T-Test.
- Placement: Dependent variable into Dependent Variables box; categorical factor into Grouping Variable box.
- Additional Controls: Activate Homogeneity test, Descriptives, and Effect Size.
- Interpretation: If , reject , indicating a statistically significant difference between group means. If Levene's test is violated, report Welch's t-test.
- Paired-Samples t-Test:
- Purpose: Compares two repeated or matched measurements collected from the same individuals.
- Navigation: Analyses T-Tests Paired Samples T-Test.
- Placement: Pair the baseline and outcome variables together.
- Interpretation: If , reject , indicating a statistically significant difference across the two conditions.
- Mann-Whitney U Test:
- Purpose: Compares rank distributions between two independent cohorts when parametric assumptions fail or data are ordinal.
- Navigation: Analyses T-Tests Independent Samples T-Test Check Mann-Whitney U under Tests.
- Interpretation: If , groups differ significantly in their mean rank distributions. It evaluates medians directly only when group distribution shapes are similar.
- Wilcoxon Signed-Rank Test:
- Purpose: Non-parametric alternative to the paired t-test; evaluates ranked differences between two related measurements.
- Navigation: Analyses T-Tests Paired Samples T-Test Check Wilcoxon signed-rank under Tests.
- Interpretation: Evaluates the ranks and signs of nonzero paired difference scores.
- One-Way ANOVA:
- Purpose: Compares the means of three or more independent groups.
- Navigation: Analyses ANOVA One-Way ANOVA.
- Placement: Dependent variable into Dependent Variable; factor into Fixed Factors.
- Controls: Check Homogeneity test and Post Hoc Tests.
- Interpretation: A significant omnibus -test () indicates at least two group means differ, but does not identify which specific pairs differ; post hoc pairwise comparisons are required.
- Kruskal-Wallis Test:
- Purpose: Compares three or more independent groups using ranks.
- Navigation: Analyses ANOVA One-Way ANOVA (Non-parametric) Kruskal-Wallis.
- Interpretation: If , at least one group's rank distribution differs significantly from the others; post hoc tests are required to locate the difference.
- Friedman Test:
- Purpose: Evaluates differences across three or more repeated or matched measurements using ranked data.
- Navigation: Analyses ANOVA Repeated Measures Non-parametric Friedman.
- Interpretation: If , performance differs significantly across repeated conditions.
- Pearson's and Spearman's Correlations:
- Purpose: Pearson evaluates linear relationships between continuous variables; Spearman evaluates monotonic relationships using ranks.
- Navigation: Analyses Regression Correlation Matrix.
- Options: Select Pearson or Spearman, along with Flag significant correlations and Confidence intervals.
Step-by-Step Test Selection Algorithm:
- Step 1: Identify Research Goal:
- Comparing groups on a variable Proceed to Step 2.
- Evaluating an association between variables Proceed to Step 3.
- Step 2: Determine Design Structure for Group Comparisons:
- Two independent cohorts: Independent-samples t-test (Parametric) or Mann-Whitney U (Non-parametric).
- Two related or paired measurements: Paired-samples t-test (Parametric) or Wilcoxon signed-rank (Non-parametric).
- Three or more independent cohorts: One-Way ANOVA (Parametric) or Kruskal-Wallis (Non-parametric).
- Three or more repeated measurements: Repeated-measures ANOVA (Parametric) or Friedman test (Non-parametric).
- Step 3: Determine the Nature of the Association:
- Linear relationship between continuous variables: Pearson correlation ().
- Monotonic relationship, ranked variables, or ordinal scales: Spearman correlation ().
- Concordance in small samples or data with tied ranks: Kendall's tau ().
- Multi-Criteria Decision Rule: Test choice should not depend on normality alone; researchers must account for research design, sample size, measurement level, presence of outliers, and the substantive hypothesis.
Statistical Formula and Concept Master Reference Sheet
Master Mathematical Formulas:
- Sample Mean:
- Range:
- Population Variance:
- Sample Variance:
- Standard Deviation:
- Relative Frequency:
- Percentage Frequency:
- Interquartile Range:
- Standardized -Score:
- Coefficient of Variation:
- Statistical Power:
- Outlier Detection Boundaries:
Core Conceptual Contrasts:
- Descriptive vs. Inferential: Summarizing sample data vs. generalizing to a broader population.
- Population vs. Sample: The complete group of interest vs. the measured subset.
- Parameter vs. Statistic: A fixed population value () vs. a sample estimate ().
- Nominal vs. Ordinal: Unordered categories vs. ordered/ranked categories.
- Interval vs. Ratio: Equal intervals without true zero vs. equal intervals with true absolute zero.
- Mean vs. Median: Mathematical center sensitive to outliers vs. positional midpoint robust to outliers.
- Variance vs. Standard Deviation: Dispersion in squared units vs. dispersion in original scale units.
- Positive vs. Negative Correlation: Variables move in the same direction vs. opposite directions.
- One-Tailed vs. Two-Tailed: Directional hypothesis with one rejection region vs. non-directional with split rejection regions.
- Type I Error () vs. Type II Error (): False positive (rejecting a true ) vs. False negative (failing to reject a false ).
Comprehensive Practice Examination and Study Guidelines
Diagnostic Practice Examination:
- Question 1: A researcher classifies students according to their college programs: Psychology, Biology, and Political Science. What is the level of measurement?
- A. Ordinal
- B. Nominal
- C. Interval
- D. Ratio
- Question 2: Which measure of central tendency is most resistant to extreme values?
- A. Mean
- B. Median
- C. Mode
- D. Range
- Question 3: A student scored at the percentile. What does this mean?
- A. The student answered correctly.
- B. The student scored higher than exactly of everyone.
- C. Approximately of the reference observations are at or below that score.
- D. The student is points above the mean.
- Question 4: Which measure describes the spread of the middle of observations?
- A. Range
- B. Variance
- C. Interquartile Range ()
- D. Mean
- Question 5: A dataset has a mean of and of . What is the -score of ?
- A.
- B.
- C.
- D.
- Question 6: In a normal distribution, approximately what percentage of observations fall within standard deviations?
- A.
- B.
- C.
- D.
- Question 7: A distribution has a long tail toward the right. What type is it?
- A. Negative skew
- B. Positive skew
- C. Symmetrical
- D. Platykurtic
- Question 8: A researcher finds with alpha = . What is the correct decision?
- A. Accept
- B. Fail to reject
- C. Reject
- D. Increase alpha
- Question 9: A researcher predicts that one group will score higher than another. Which test direction is appropriate if justified in advance?
- A. Two-tailed
- B. One-tailed
- C. No-tailed
- D. Non-parametric only
- Question 10: A researcher asks whether two variables are related without predicting positive or negative association. What is the hypothesis?
- A. Directional
- B. Non-directional
- C. Null only
- D. Causal
- Question 11: A researcher rejects a null hypothesis that is actually true. What error occurred?
- A. Type I
- B. Type II
- C. Sampling
- D. Measurement
- Question 12: A researcher fails to reject a false null hypothesis. What error occurred?
- A. Type I
- B. Type II
- C. No error
- D. Calculation error
- Question 13: Which test is used to examine whether the data deviate significantly from normality?
- A. Levene's test
- B. Shapiro-Wilk test
- C. Pearson correlation
- D. Mann-Whitney U
- Question 14: Levene's test produces in an independent-samples comparison. What does this suggest?
- A. Equal variances
- B. Unequal variances
- C. Normal distribution
- D. Strong correlation
- Question 15: Which test compares the means of two independent groups?
- A. Paired t-test
- B. Independent-samples t-test
- C. Friedman test
- D. Pearson correlation
- Question 16: The same students are measured before and after an intervention. Which parametric test is appropriate?
- A. Independent t-test
- B. Paired t-test
- C. One-way ANOVA
- D. Kruskal-Wallis
- Question 17: Which test is commonly used as a non-parametric alternative to the independent-samples t-test?
- A. Wilcoxon signed-rank
- B. Mann-Whitney U
- C. Friedman
- D. Pearson
- Question 18: A researcher compares three independent groups using a non-parametric test. Which should be considered?
- A. Kruskal-Wallis
- B. Wilcoxon signed-rank
- C. Paired t-test
- D. Spearman
- Question 19: A researcher wants to measure a linear relationship between two quantitative variables. Which test is most appropriate when assumptions are met?
- A. Spearman
- B. Pearson
- C. Mann-Whitney
- D. Friedman
- Question 20: A correlation coefficient is . What does this indicate?
- A. Weak positive relationship
- B. Strong negative relationship
- C. No relationship
- D. Strong positive relationship
Answer Key with Explanations:
- Question 1: B (Nominal classifies subjects into categories without order).
- Question 2: B (Median is robust against extreme values).
- Question 3: C (Percentiles indicate the percentage of observations at or below a score).
- Question 4: C (Interquartile Range spans the middle of observations, ).
- Question 5: C ().
- Question 6: C (Empirical rule defines as covering approximately ).
- Question 7: B (A tail pointing right indicates a positive skew).
- Question 8: C (Since , the decision is to reject ).
- Question 9: B (Predicting a directional outcome justifies a one-tailed test).
- Question 10: B (Hypothesizing an unspecified relationship constitutes a non-directional hypothesis).
- Question 11: A (Rejecting a true null hypothesis is a Type I error).
- Question 12: B (Failing to reject a false null hypothesis is a Type II error).
- Question 13: B (Shapiro-Wilk tests for significant departures from normality).
- Question 14: B (Levene's indicates significant variance inequality).
- Question 15: B (Independent-samples t-test compares means between two separate cohorts).
- Question 16: B (A pre/post design with the same subjects uses a paired t-test).
- Question 17: B (Mann-Whitney U is the non-parametric counterpart to the independent t-test).
- Question 18: A (Kruskal-Wallis compares three or more independent groups non-parametrically).
- Question 19: B (Pearson's measures linear relationships between continuous variables).
- Question 20: B ( reflects a strong negative linear association).
Priority Study Areas and Strategic Guidance:
- Five Core Focus Areas:
- Variables and Measurement: Differentiating Nominal, Ordinal, Interval, and Ratio scales.
- Descriptive Statistics: Computing and interpreting the Mean, Median, Mode, Range, Variance, and Standard Deviation.
- Hypothesis Testing Framework: Mastering -value decision rules, one-tailed vs. two-tailed tests, directional vs. non-directional hypotheses, and rejection criteria.
- Errors and Assumptions: Understanding Type I () and Type II () errors, and interpreting the Shapiro-Wilk and Levene's tests.
- Statistical Test Selection: Identifying the correct parametric or non-parametric test based on research design and measurement scale.
- Foundational Decision Questions:
- Am I evaluating differences between groups or examining an association between variables?
- How many groups or repeated measurement points are present?
- Are the observations independent or related/paired?
- What level of measurement characterizes the variables?
- Are parametric distributional assumptions met, or are non-parametric rank tests required?