Statistics for the Behavioral Sciences

Chapter 1: Introduction and Descriptive Statistics

The Use of Statistics in Science

Statistics is defined as the analysis and evaluation of scientific observations. It provides a structured mathematical framework for quantifying data and drawing meaningful conclusions. The field is broadly categorized into two main branches:

• Descriptive Statistics: Used to summarize and describe the characteristics of a specific data set. It is the first step in statistical analysis, helping to make large volumes of data more manageable.

• Inferential Statistics: Used to make generalizations or inferences about a larger population based on the data collected from a sample.

Descriptive and Inferential Statistics

Terminology: Data vs. Datum

• Datum: A single measurement or observation, often referred to as a score or raw score.

• Data: Multiple measurements or observations collected together.

Presentation of Data

Descriptive statistics involve the use of tables, graphs, and summary measures (like the mean or standard deviation). For example, making scientific observations requires a systematic way to present findings so patterns become apparent.

Populations and Samples

• Population: The entire set of individuals, items, or data points of interest in a study.

• Population Parameter: A characteristic (usually numeric) that describes the entire population.

• Sample: A subset of individuals selected from the population, intended to represent the larger group.

• Sample Statistic: A characteristic that describes the sample data. Statistics are used to estimate parameters.

Research Methods and Statistics

Data collection must follow strict rules to ensure validity. The Scientific Method, or Research Method, is the systematic process of gathering information.

1. Experimental Method

This method is designed to demonstrate cause-and-effect relationships by controlling conditions. Key requirements include:

• Manipulation: The researcher actively changes the level of the Independent Variable (IV). The various levels of the IV are the conditions being compared.

• Randomization (Random Assignment): Participants have an equal chance of being assigned to any group, which helps control for preexisting differences.

• Comparison/Control: Including a control group (no treatment) to compare against treatment groups.

• Dependent Variable (DV): The variable that is measured; it is hypothesized to be affected by the IV.

• Operational Definition: A clear description of how a variable is measured or defined within a study (e.g., defining "fear" as a specific heart rate threshold).

2. Quasi-Experimental Method

Used when researchers test a hypothesis but lack full control over conditions.

• Quasi-independent variable (Factor): A variable that differentiates groups (like gender or age) that cannot be randomly assigned.

• This method lacks the random assignment found in true experiments, and there may not be a true comparison or control group.

3. Correlational Method

This method measures the extent to which two variables change together.

• It involves no control over experimental conditions.

• Researchers observe variables in their natural state to determine if a relationship (correlation) exists between them.

Scales of Measurement

Measurements are characterized by four properties: identity, order, difference, and ratio.

• Nominal Scale: Represents an identity or category. It provides no information about rank or quantity. Coding is often used here (e.g., assigning 1 for "Male" and 2 for "Female").

• Ordinal Scale: Conveys rank or order. It shows that one value is greater or less than another, but the distance between ranks is not equal.

• Interval Scale: Features an equidistant scale where the difference between values is consistent. It lacks a True Zero (e.g., Temperature in Fahrenheit).

• Ratio Scale: The most informative scale. It has equidistant points and a true zero point, representing the total absence of the variable (e.g., height, weight, or time).

Types of Variables

• Continuous Variable: Can take on an infinite number of values between any two points (e.g., distance).

• Discrete Variable: Measured in whole units or categories; there are no values between adjacent units (e.g., number of children).

• Quantitative Variable: Data that represents amounts or counts.

• Qualitative Variable: Data that describes labels, categories, or qualities.

SPSS in Focus: Entering and Defining Variables

SPSS (Statistical Package for the Social Sciences) is a common tool for data analysis.

• Data Entry: Scores (such as GPA) are typically entered into a spreadsheet.

• Columns: Usually represent variables.

• Rows: Usually represent individual cases or subjects.

• A final check of the data entry is essential for accuracy.

Chapter 2: Frequency Distributions in Tables and Graphs

Why Summarize Data?

Summarizing data allows for comparing scores and identifying the Frequency (the number of times a score occurs). This is especially critical for large data sets to make them meaningful and interpretable.

Simple Frequency Distributions for Grouped Data

A Simple Frequency Distribution summarizes the frequency of scores in categories. When there are many scores, Grouped Data is used to organize scores into Intervals.

Steps to Construct a Distribution:

1. Find the Real Range: The difference between the highest and lowest scores in the data set.

2. Find the Interval Width ($i$):     ${Interval Width = \frac{Real \, range}{Number \, of \, intervals}}$

3. Construct the Frequency Distribution: List intervals and count the frequency of scores within each.

Rules for Grouped Data:

• Intervals must be mutually exclusive (no overlapping boundaries).

• All intervals must be the same width.

• Open Interval: An interval with no specific upper or lower boundary (e.g., "90 and above").

• Outlier: An extreme score that falls significantly outside the range of most of the data.

Other Ways of Summarizing Grouped Data

• Cumulative Frequency ($cf$): The sum of frequencies at and below a particular interval (bottom-up) or at and above (top-down).

• Relative Frequency ($rf$): The proportion of total scores in an interval.     ${Relative \, Frequency = \frac{Observed \, frequency}{Total \, frequency \, count}}$

• Relative Percent: The percentage of the total data set represented by an interval.     ${Relative \, Percent = \frac{Observed \, frequency}{Total \, frequency \, count} \times 100}$

• Cumulative Relative Frequency / Cumulative Percent: The sum of relative frequencies or percents up to a certain point.

Percentile Points and Percentile Ranks

• Percentile Point: The raw score value at or below which a specific percentage of scores fall.

• Percentile Rank: The percentage of scores in a distribution that are at or below a given raw score.

Graphs for Continuous Data

Graphs provide an alternative visual way to display frequency distributions.

• Histogram: A bar-style graph where bars touch, used for continuous data. The height represents frequency.

• Frequency Polygon: A dot-and-line graph where dots are placed at the midpoint of each interval and connected by lines.

• Ogive: A line graph representing cumulative frequencies.

• Stem-and-Leaf Display: A display that lists individual scores. The Stem (left of the vertical line) represents the leading digits, and the Leaf (right of the line) represents the trailing digits. It retains the identity of every single score, unlike a histogram.

Graphs for Discrete and Categorical Data

• Bar Chart: Similar to a histogram, but the bars do not touch, emphasizing the discrete nature of the categories.

• Pie Chart: A circular graph where the "slices" or Sectors represent the proportion of the total for each category.

Chapter 3: Central Tendency

Introduction to Central Tendency

Central tendency describes the center of a distribution. Measures of central tendency for samples are Sample Statistics, while those for populations are Population Parameters ($N$ vs. $n$).

Measures of Central Tendency

• Mean ($M$ or $\mu$): The arithmetic average.     - Weighted Mean: Used when combining groups with unequal sample sizes.     ${Weighted \, mean = \frac{Sum \, of \, weighted \, products}{Sum \, of \, weights}}$

• Median: The middle score in a distribution when ordered from lowest to highest. It is at the ${(n+1)/2}$ position. It is unaffected by outliers.

• Mode: The value that occurs most frequently. A distribution can be unimodal, bimodal, multimodal, or nonmodal (rectangular).

Characteristics of the Mean

• Every score in the distribution affects the mean.

• Adding/removing a score usually changes the mean (unless the score equals the mean).

• Adding, subtracting, multiplying, or dividing every score by a Constant changes the mean by that same constant.

• The sum of the differences of each score from the mean is always zero: ${ \sum (x - M) = 0}$. This makes the mean the Balancing Point of the distribution.

• The sum of the squared differences of scores from the mean ${ \sum (x - M)^2 }$ is minimal compared to any other value.

Choosing a Measure

• Mean: Best for Normal Distributions and Interval/Ratio data.

• Median: Best for Skewed Distributions (where outliers pull the mean) and Ordinal data.

• Mode: Best for Nominal data and describing Modal Distributions.

Chapter 4: Summarizing Data: Variability

Introduction to Variability

Variability measures the spread or dispersion of scores in a distribution. Knowing the mean is not enough; variability explains how far scores deviate from that center.

Basic Measures

• Range: The difference between the largest and smallest values. It is sensitive to outliers and is often avoided by researchers.

• Quartiles: Fractiles that divide the data into four equal parts.     - Interquartile Range (IQR): ${Q3 - Q1}$.     - Semi-interquartile Range (SIQR): ${(Q3 - Q1) / 2}$, which describes the spread of the middle 50% of scores.

Variance

Variance measures the average squared distance from the mean.

• Population Variance ($\sigma^2$): ${ \sigma^2 = \frac{\sum (X - \mu)^2}{N} }$

• Sample Variance ($s^2$): Uses ${n - 1}$ in the denominator to be an Unbiased Estimator of the population variance.     ${ s^2 = \frac{\sum (X - M)^2}{n - 1} = \frac{SS}{df} }$

• Sums of Squares (SS): The numerator of the variance formula (${\sum (X - M)^2}$).

Standard Deviation

The standard deviation is the square root of the variance, bringing the unit of measurement back to the original scale.

• Population ($\sigma$): ${ \sigma = \sqrt{\sigma^2} }$

• Sample ($SD$): ${ SD = \sqrt{s^2} }$

Informativeness of Standard Deviation

• Empirical Rule: In a normal distribution, approximately 68% of scores fall within $1$ SD, 95% within $2$ SD, and 99.7% within $3$ SD of the mean.

• Chebyshev’s Theorem: For any distribution, the proportion of scores within $k$ standard deviations is at least $1-\frac{1}{k^{2^{\prime}}}$ . (Applies to non-normal distributions).

• Standard Deviation Characteristics: Always positive, affected by every score, reported as $M\pm SD$

Chapter 5: Probability and the Foundations of Inferential Statistics

Introduction to Probability

Probability predicts the likelihood of random events. It ranges from $0$ to $1$ and can never be negative. ${ P(A) = \frac{Count(A)}{Sample \, space} }$

Relationships Between Outcomes

• Mutually Exclusive: Outcomes that cannot occur at the same time. Use the Additive Rule: ${ p(A \cup B) = p(A) + p(B) }$.

• Independent: The occurrence of one outcome does not affect the other. Use the Multiplicative Law: ${ p(A \cap B) = p(A) \times p(B) }$.

• Complementary: The sum of probabilities equals 1. ${ p(A) = 1 - p(B) }$.

• Conditional: One outcome depends on another. ${ p(U|P) = \frac{p(P \cap U)}{p(P)} }$.

• Bayes’s Theorem: ${ p(U|P) = \frac{p(P|U)p(U)}{p(P)} }$.

Probability Distributions

• Random Variable: A variable whose values are determined by chance.

• Expected Value: The mean of a probability distribution. { \mu = \sum [x \cdot p(x)] }$.

• Variance of Probability Distribution: { \sigma^2 = \sum [(x - \mu)^2 \cdot p(x)] = \sum (x^2 \cdot p(x)) - \mu^2 }$.</p></li></ul><h5 id="dc59371a-d437-44f2-b48a-cdd2c52a0139" data-toc-id="dc59371a-d437-44f2-b48a-cdd2c52a0139" collapsed="false" seolevelmigrated="true">Binomial Distribution</h5><p>A distribution for trials with only two outcomes (Success/Failure).</p><ul><li><p><strong>Mean</strong>:${ \mu = np }$

• Variance: ${ \sigma^2 = npq }$ (where ${q = 1 - p}$).

• Standard Deviation: { \sigma = \sqrt{npq} }$

Chapter 6: Probability, Normal Distributions, and z Scores

Characteristics of the Normal Distribution

Mathematically defined by de Moivre, it is theoretical, symmetrical, and unimodal.

• Mean, median, and mode are equal.

• Area under the curve = 1$.</p></li><li><p>Tails are <strong>Asymptotic</strong> (never touch the X-axis).</p></li><li><p>Standard deviation is always positive.</p></li></ul><h5 id="d5dd3b6b-e0ed-4fa9-b843-9c6bee85ae57" data-toc-id="d5dd3b6b-e0ed-4fa9-b843-9c6bee85ae57" collapsed="false" seolevelmigrated="true">The z-Transformation</h5><p>Converts a raw score ($X$) into standard units (number of standard deviations from the mean).</p><ul><li><p><strong>Population</strong>:${ z = \frac{X - \mu}{\sigma} }$</p></li><li><p><strong>Sample</strong>:${ z = \frac{X - M}{SD} }$</p></li></ul><h5 id="893ed7bf-72c4-4d6f-9873-42053f4683cd" data-toc-id="893ed7bf-72c4-4d6f-9873-42053f4683cd" collapsed="false" seolevelmigrated="true">The Unit Normal Table</h5><ul><li><p><strong>Column A</strong>: Lists z-scores.</p></li><li><p><strong>Column B</strong>: Area between the mean and the z-score.</p></li><li><p><strong>Column C</strong>: Area from the z-score toward the tail.</p></li></ul><h5 id="bc1ca5b9-81af-4fcf-9160-b3d5c22e179f" data-toc-id="bc1ca5b9-81af-4fcf-9160-b3d5c22e179f" collapsed="false" seolevelmigrated="true">Normal Approximation to the Binomial</h5><p>As sample size ($n$) increases, the discrete binomial distribution approximates the continuous normal distribution. Conditions must be checked (usually${np \geq 10}$and${nq \geq 10}$).</p><ul><li><p>Requires finding <strong>Real Limits</strong> for discrete values before calculating z.</p></li></ul><h4 id="fc5e86fb-4fa8-4945-8454-7676e77e7700" data-toc-id="fc5e86fb-4fa8-4945-8454-7676e77e7700" collapsed="false" seolevelmigrated="true">Chapter 7: Probability and Sampling Distributions</h4><h5 id="19c62342-09ae-4b9a-85b3-9e3bded148d0" data-toc-id="19c62342-09ae-4b9a-85b3-9e3bded148d0" collapsed="false" seolevelmigrated="true">Selecting Samples</h5><ul><li><p><strong>Sampling with Replacement</strong>: Member is returned to the population before the next draw (keeps probabilities constant).</p></li><li><p><strong>Sampling without Replacement</strong>: Member is not returned (changes probabilities).</p></li><li><p><strong>Theoretical Sampling</strong>: Uses every possible sample of size$n$from a population$N$. Total samples =$N^{n}$</p></li><li><p><strong>Experimental Sampling</strong>: Used in behavioral research. Total samples =$\frac{N!}{n!\left(N-n\right)!}$</p></li></ul><h5 id="6fe9ccae-7416-4851-9c75-cd8e2b0aa0b2" data-toc-id="6fe9ccae-7416-4851-9c75-cd8e2b0aa0b2" collapsed="false" seolevelmigrated="true">Central Limit Theorem</h5><p>Regardless of the shape of the population distribution, the sampling distribution of the mean will become normal as sample size ($n$) increases.</p><ul><li><p><strong>Unbiased Estimator</strong>: The mean of the sampling distribution of the mean ($M_M$) is equal to the population mean ($\mu$).</p></li></ul><h5 id="f8f2a1e8-12f5-4e18-baea-f263be0f0751" data-toc-id="f8f2a1e8-12f5-4e18-baea-f263be0f0751" collapsed="false" seolevelmigrated="true">Standard Error of the Mean ($s_M$)</h5><p>The standard deviation of the sampling distribution. It measures the <strong>Sampling Error</strong> (discrepancy between sample and population).${ \sigma_M = \frac{\sigma}{\sqrt{n}} }$</p><ul><li><p><span>Decreased by a smaller population SD () or a larger sample size () (</span><strong><span>Law of Large Numbers</span></strong><span>).</span></p></li></ul><h4 id="f25f069d-e532-4305-8e9e-fff21fe37591" data-toc-id="f25f069d-e532-4305-8e9e-fff21fe37591" collapsed="false" seolevelmigrated="true">Chapter 8: Hypothesis Testing: Significance, Effect Size, Estimation, and Power</h4><h5 id="f3a410e7-ce93-4851-803c-ec56b213c8e9" data-toc-id="f3a410e7-ce93-4851-803c-ec56b213c8e9" collapsed="false" seolevelmigrated="true">The Inferential Step</h5><p>We observe samples to learn about population effects. <strong>Null Hypothesis Significance Testing (NHST)</strong> follows four steps:</p><ol><li><p><strong>State the Hypotheses</strong>: Null (${H_0}$) usually predicts no effect. Alternative (${H_1}) predicts an effect.

• Set the Criteria: Select Alpha ($\alpha$), the probability of a Type I error (usually 0.05$).</p></li><li><p><strong>Compute Test Statistic</strong>: (e.g., z-test, t-test).</p></li><li><p><strong>Make a Decision</strong>: If${p < \alpha}$, reject${H_0}. The result is Statistically Significant.

Types of Errors

• Type I Error ($\alpha$): Rejecting {H_0} when it is actually true (False Positive).

• Type II Error ($\beta$): Failing to reject {H_0}$when it is actually false (False Negative).</p></li><li><p><strong>Type III Error</strong>: Making a directional prediction, but the effect is in the opposite direction.</p></li><li><p><strong>Power ($1 - \beta$)</strong>: The probability of correctly rejecting a false null hypothesis.</p></li></ul><h5 id="0d300e40-20c2-441a-b7f3-bab19126b18f" data-toc-id="0d300e40-20c2-441a-b7f3-bab19126b18f" collapsed="false" seolevelmigrated="true">Testing Significance: The z-Test</h5><p>Used when the population means and variance are known.${ z_{obt} = \frac{M - \mu}{\sigma_M} }$</p><ul><li><p><strong>Directional (One-tailed)</strong>:${H_1: > \mu}$or${H_1: < \mu}$.</p></li><li><p><strong>Nondirectional (Two-tailed)</strong>:${H_1: \neq \mu}$.</p></li></ul><h5 id="49765ef7-7f3c-4d54-9044-a7026322da71" data-toc-id="49765ef7-7f3c-4d54-9044-a7026322da71" collapsed="false" seolevelmigrated="true">Effect Size: Cohen’s d</h5><p>Hypothesis tests only tell if an effect exists; effect size tells how large it is.${ d = \frac{M - \mu}{\sigma} }$</p><ul><li><p>Cohen's Conventions: Small ($0.2$), Medium ($0.5$), Large ($0.8$).</p></li></ul><h5 id="a2c3c9f9-9335-4a56-84d3-4ba9e538d650" data-toc-id="a2c3c9f9-9335-4a56-84d3-4ba9e538d650" collapsed="false" seolevelmigrated="true">Confidence Intervals (CI)</h5><p>A range of values that likely contains the population means.${ CI = M \pm z_{crit}(\sigma_M) }$</p><h4 id="c3344392-100a-4e46-90b2-89e1a7210dfd" data-toc-id="c3344392-100a-4e46-90b2-89e1a7210dfd" collapsed="false" seolevelmigrated="true">Chapter 9: Testing Means: One-Sample t Test</h4><h5 id="2b526bf4-6871-43c9-86e6-2ac309a0d8ac" data-toc-id="2b526bf4-6871-43c9-86e6-2ac309a0d8ac" collapsed="false" seolevelmigrated="true">Going from z to t</h5><p>If the population variance is unknown, we use sample variance to compute the <strong>Estimated Standard Error</strong> ($s_M$).${ s_M = \frac{SD}{\sqrt{n}} }$${ t_{obt} = \frac{M - \mu}{s_M} }$</p><h5 id="98844156-d474-4fc4-a81b-44f8fa5e61ce" data-toc-id="98844156-d474-4fc4-a81b-44f8fa5e61ce" collapsed="false" seolevelmigrated="true">Degrees of Freedom ($df$)</h5><p>For a one-sample t-test,${ df = n - 1 }$. The t-distribution changes shape based on${df}$, becoming more normal as${df}$increases.</p><h5 id="cf32b2ac-ab9c-40c0-9963-79dfddb9f162" data-toc-id="cf32b2ac-ab9c-40c0-9963-79dfddb9f162" collapsed="false" seolevelmigrated="true">Effect Size for t-Test</h5><ul><li><p><strong>Estimated Cohen’s d</strong>: <span style="font-family: Arial;"><span>a measure of effect size in terms of the number of standard deviations that mean scores shift above or below the population mean stated by the null hypothesis.$d=\frac{M-\mu}{SD}

• Eta-Squared: Proportion of variance explained by the treatment.     { \eta^2 = \frac{t^2}{t^2 + df} }

• Omega-Squared: A more conservative estimate than eta-squared.     { \omega^2 = \frac{t^2 - 1}{t^2 + df} }$</p></li></ul><h4 id="d4a704d7-7c7e-459b-b766-5bea6e0e32b6" data-toc-id="d4a704d7-7c7e-459b-b766-5bea6e0e32b6" collapsed="false" seolevelmigrated="true">Chapter 10: Two-Independent-Sample t Test</h4><h5 id="2693a01d-0a22-4f49-89b7-768ebca0494a" data-toc-id="2693a01d-0a22-4f49-89b7-768ebca0494a" collapsed="false" seolevelmigrated="true">Between-Subjects Design</h5><p>Compares two independent samples (different people in each group).</p><ul><li><p><strong>Pooled Variance ($s_p^2): Weighted average of the variances from both samples.

• Estimated Standard Error:     { s_{M1 - M2} = \sqrt{\frac{s_p^2}{n1} + \frac{s_p^2}{n2}} }$${ t_{obt} = \frac{(M1 - M2) - (\mu 1 - \mu 2)}{s_{M1-M2}} }$</p></li></ul><h5 id="53d5fbed-828c-422f-b894-309dd700c802" data-toc-id="53d5fbed-828c-422f-b894-309dd700c802" collapsed="false" seolevelmigrated="true">Assumptions</h5><ol><li><p><strong>Normality</strong>: Scores are normally distributed.</p></li><li><p><strong>Homogeneity of Variance</strong>: Variances are equal across groups (checked via Levene’s test in SPSS).</p></li><li><p><strong>Independence</strong>: Individual scores are independent.</p></li></ol><h4 id="ca24cf57-18c2-4ac6-aa82-3a1ee6f6a748" data-toc-id="ca24cf57-18c2-4ac6-aa82-3a1ee6f6a748" collapsed="false" seolevelmigrated="true">Chapter 11: Related-Samples t Test</h4><h5 id="f83ee014-0f85-4d7a-8c59-c629a36f40cd" data-toc-id="f83ee014-0f85-4d7a-8c59-c629a36f40cd" collapsed="false" seolevelmigrated="true">Study Designs</h5><ul><li><p><strong>Repeated-Measures</strong>: Same participants observed in both levels (Within-subjects).</p></li><li><p><strong>Matched-Pairs</strong>: Participants are paired based on shared traits (e.g., twins or similar IQ).</p></li><li><p><strong>Advantages</strong>: More practical, reduces standard error by eliminating between-persons error, and increases power.</p></li></ul><h5 id="76908dea-7eb1-4922-80e1-f854349202f2" data-toc-id="76908dea-7eb1-4922-80e1-f854349202f2" collapsed="false" seolevelmigrated="true">Calculations</h5><p>Based on <strong>Difference Scores ($D$)</strong> (${D = X_1 - X_2}$).</p><h4 id="a83b5437-59d8-4f5b-ab87-332ae35f764e" data-toc-id="a83b5437-59d8-4f5b-ab87-332ae35f764e" collapsed="false" seolevelmigrated="true">Chapter 12: Analysis of Variance (ANOVA): One-Way Between-Subjects</h4><h5 id="77915e9f-fe37-45d4-9c4f-000f6c0cce74" data-toc-id="77915e9f-fe37-45d4-9c4f-000f6c0cce74" collapsed="false" seolevelmigrated="true">Why ANOVA?</h5><p>Used when comparing more than two groups to avoid "experimentwise" alpha inflation (multiple t-tests would increase the risk of Type I error).</p><h5 id="8dcf8ecc-9801-42bf-af5a-887941b00b36" data-toc-id="8dcf8ecc-9801-42bf-af5a-887941b00b36" collapsed="false" seolevelmigrated="true">Sources of Variation</h5><ul><li><p><strong>Between-Groups Variation</strong>: Variance attributed to the treatment or factor.</p></li><li><p><strong>Within-Groups Variation (Error)</strong>: Unexplained individual differences.</p></li><li><p><strong>F-ratio</strong>:${ F = \frac{MS_{BG}}{MS_E} }$. If it is true.</p></li><li><p><strong>Mean Square (MS)</strong>:${ MS = \frac{SS}{df} }$.

Post Hoc Tests

If F is significant, post hoc tests determine which specific pairs differ.

• Tukey’s HSD (Honestly Significant Difference): Most common.

• Scheffe and Bonferroni are other alternatives.

Chapter 13: One-Way Within-Subjects ANOVA

Characteristics

Used for repeated-measures designs.

• Sources of Variation: Total variability is split into Between-Groups, Between-Persons, and Error.

• Sphericity Assumption: Equal variance of the differences between all level combinations (checked by Mauchly’s test).

• Bonferroni Procedure: Often used here to adjust alpha levels for post hoc comparisons.

Effect Size

• Partial Eta-Squared ($\eta_p^2$ ): $\eta_{p}^2=\frac{SS_{BG}}{SS_{T}-SS_{BP}};\eta_{p}^2=\frac{SS_{BG}}{SS_{BG}+SS_{E}}$

• This removes between-persons variance from the calculation.

Chapter 14: Two-Way Between-Subjects Factorial Design

Complexity and Terminology

• Factorial Design: Involves two or more factors.

• Main Effect: The effect of one factor regardless of the other.

• Interaction: Occurs when the effect of one factor depends on the level of the other factor (seen as non-parallel lines on a graph).

Analysis

A two-way ANOVA provides three hypothesis tests: Main Effect A, Main Effect B, and the ${A \times B}$ Interaction.

• Simple Main Effect Test: Conducted if an interaction is significant, testing the effect of one factor at a single level of the other factor.

Chapter 15: Correlation

Measuring Relationships

• Correlation Coefficient ($r$): Measures the direction (+/-) and strength ($0$ to $1$) of a relationship.

• Pearson Correlation: Used for interval/ratio data.

• Spearman Correlation ($r_s$): Used for ordinal/ranked data. $r_{s}=1-\frac{_6\Sigma_{D^2}}{n\left(n^2-1\right)}$

• Point-Biserial ($r_{pb}$): One continuous and one dichotomous variable.

• Phi Coefficient ($\phi$): Two dichotomous variables.

Limitations

• Causality: Correlation does not imply causation.

• Outliers: Can significantly obscure or exaggerate relationships.

• Restriction of Range: Narrow data sets can hide true relationships.

Chapter 16: Linear Regression and Multiple Regression

Linear Regression

Uses the relationship from correlation to predict values.

• Regression Line: The "best-fitting" line (${ Y = bX + a }$).

• Slope ($b$) and y-intercept ($a$).

• Method of Least Squares: Minimizes the sum of squared residuals (${ SS_{residual} }$).

• Standard Error of Estimate ($s_e$): Measures average prediction error.

Multiple Regression

Predicts a criterion variable ($Y$) using multiple predictor variables ($X_1, X_2, …$).

• Multicollinearity: A problem where predictors are too highly correlated with each other (checked via VIF).

• Beta Coefficient ($\beta$): Standardized coefficients that show the unique relative contribution of each predictor.

Chapter 17: Nonparametric Tests: Chi-Square Tests

Introduction to Chi-Square (\chi^2)

Chi-Square tests are essential statistical tools used primarily to analyze categorical data, particularly for evaluating relationships between variables or testing hypotheses about observed frequencies across categories. Unlike parametric tests, which assume the data follow a specific distribution, Chi-Square tests do not require such assumptions, making them particularly useful in research contexts where the data are nominal or ordinal.

Goodness-of-Fit Test

This test assesses how well the observed frequency distribution of a categorical variable aligns with an expected distribution derived from a theoretical hypothesis.
 

• Purpose: To ascertain whether a sample distribution fits a specified distribution.

• Formula: The test statistic is calculated using the formula:   $\chi^2_{obt} = \sum \frac{(f_o - f_e)^2}{f_e}$
    where:   - f_o: Observed frequency for each category.
    - f_e: Expected frequency under the null hypothesis.

• Interpretation: A significant Chi-Square value indicates that the observed frequencies differ markedly from the expected frequencies, leading to rejection of the null hypothesis that there is no difference between the observed and expected frequencies.

Test for Independence

This test evaluates whether two categorical variables are independent, allowing researchers to determine whether a relationship exists.
 

• Purpose: To explore the association between two categorical variables, often assessed in contingency tables.

• Procedure: The test computes a Chi-Square statistic based on the frequencies observed in a cross-tabulation format, where f_o is the observed frequency from the table, and f_e is calculated based on the products of the row and column totals divided by the grand total.

• Interpretation: A significant result indicates that the two variables are likely related, while a non-significant result suggests independence.

Assumptions of Chi-Square Tests

To properly utilize Chi-Square tests, certain assumptions must be met:

• Independence: Each observation should be independent of others. This ensures that the data points are not influenced by one another.

• Minimum Frequency: Each expected frequency in the contingency table should be at least 5. If this condition is not met, the test results may be unreliable.

• Categorical Data: The data should be nominal or ordinal, as Chi-Square tests cannot be used for continuous data without first converting it to categories.

Cramer’s V

Cramer’s V is a measure of effect size associated with the Chi-Square test for independence. It quantifies the strength of association between two categorical variables, providing insight into the practical significance of the observed relationship.

• Calculation: It is calculated as follows:   $V = \sqrt{\frac{\chi^2}{n \cdot \min(k-1,r-1)}}$   where:   - n = total number of observations   - k = number of columns   - r = number of rows

• Interpretation: Values range from 0 to 1. A value closer to 0 suggests a weak association, while a value closer to 1 indicates a strong association.

Applications of Chi-Square Tests

Chi-Square tests are broadly applicable across various disciplines such as:

• Social Sciences: Examining the relationship between responses to survey items.

• Healthcare: Investigating the association between lifestyle factors and health outcomes.

• Market Research: Exploring consumer preferences across different demographic groups.

• Psychology: Analyzing frequencies of categorical responses to psychological assessments.

Limitations of Chi-Square Tests

Despite their utility, Chi-Square tests have limitations that researchers must consider:

• Sensitivity to Sample Size: Large sample sizes can inflate Chi-Square statistics, leading to significant results that may not be practically significant.

• Data Type Restrictions: They can only be used with categorical data and are not suited for continuous variables without proper categorization.

• Loss of Information: Grouping continuous data into categories can lead to information loss, potentially affecting the results.

Conclusion

Chi-Square tests remain fundamental tools in statistical analysis for categorical data. Their flexibility and nonparametric nature make them accessible for researchers examining relationships in various fields, though careful attention to assumptions and limitations is crucial for valid and reliable results.