Statistics in Psychology — Comprehensive Study Notes (Unit 1)

Population vs. Sample

Population: all people of interest for a study. Example: all college students in the United States or all students at Saint Philip's College (SPC).
Sample: a subset drawn from the population for study. Must be a good representation of the population to generalize results.
Why we use samples: realistic studies cannot test every member of a population (budgets, time, practicality).
Three basic criteria for a good sample:
- Representativeness: includes the diversity of the population (e.g., genders, races/ethnicities, ages, SES).
- Randomness: every member of the population has an equal chance of being selected; avoid handpicking to avoid bias.
- Adequate size: large enough to generalize, not too small to avoid skew from outliers.
Example: sampling US college students where 97 out of 100 are female would not be representative of a population that includes multiple genders; representativeness is key for generalization.

Data, Variables, and Measurements

Data: information gathered from observations or measurements in a study.
Variable: any characteristic that differs between individuals; can be manipulated (independent) or measured (dependent).
Operational definition: how a variable is defined for measurement; converts abstract concepts (e.g., hunger, aggression) into measurable quantities.
Data set: the collection of measurements or observations from a study.
Measurements vs scores: individual measurements = scores; the entire collection = data set.
Relationship to methods: variables and measurements underpin how we apply statistical methods.

Descriptive vs. Inferential Statistics

Descriptive statistics: summarize, organize, and describe data (e.g., mean, median, mode, range, standard deviation).
Inferential statistics: allow us to draw generalizations about populations from samples; address sampling error and uncertainty.
Core goal: extend findings from a sample to a population with quantified uncertainty.
Core problem: sampling error—the discrepancy between a sample statistic and the population parameter.
Key concepts to learn later (as mentioned in lecture): normal distribution, z scores, correlation vs. causation, significance, data sets, ordinal vs. nominal variables, etc.

Population Parameters vs. Sample Statistics

Population parameter: a numerical value that describes a population (e.g., mean μ, standard deviation σ).
Sample statistic: a numerical value calculated from a sample (e.g., sample mean ar{x}, sample standard deviation $s$ ).
Relationship: sample statistics are used to estimate population parameters.
Why this matters: because we cannot measure every member of a population, we rely on sample statistics to infer population values.
Example: if the average income of all investment bankers in NYC is μ, but we can only sample 500 bankers and get a sample mean ar{x}, we use ar{x} to estimate μ.
Inference from samples relies on representativeness and size to minimize error.

Sampling Error and Estimation

Sampling error: the natural discrepancy between a sample statistic and the population parameter it estimates.
No sample is perfect; different samples yield different statistics.
To reduce sampling error, we use well-designed samples and statistical methods (hypothesis testing, confidence intervals, etc.).
Standard error of the mean (illustrative):
- Population level: SE_{ar{X}} = \frac{\sigma}{\sqrt{n}}
- Sample level (when σ is unknown): SE_{ar{X}} = \frac{s}{\sqrt{n}}
Repeated sampling idea: take many random samples and look at the distribution of their statistics; this helps us estimate population parameters more accurately.
Concept of power and sample size: larger samples generally reduce sampling error and increase the likelihood of detecting true effects (statistical power).

Descriptive Statistics in Practice

Common descriptive measures:
- Mean: $\bar{x} = \frac{1}{n} \sum{i=1}^n xi$
- Median: middle value when data are ordered
- Mode: most frequent value
- Range: $R = x{\max} - x{\min}$
- Standard deviation: $s = \sqrt{\frac{1}{n-1} \sum{i=1}^n (xi - \bar{x})^2}$
Use cases: summarize thousands or millions of scores; identify outliers; observe distribution patterns.
Why descriptive stats alone are not enough: to answer questions about relationships and effects, we need inferential methods.
Preview of future topics: normal distribution, z-scores, variables, correlation vs. causation, and significance.

Inferential Statistics: Making Inferences from Samples

Inferential statistics: methods that let us study samples and generalize to populations.
Core problem they address: sampling error and how to quantify confidence in population estimates.
Key questions addressed: Is observed difference due to manipulation or chance? How confident are we in our inference?
Hypothesis testing and confidence intervals are tools to address these questions (to be covered in depth later).

Correlational Studies vs. Experimental Studies

Correlational studies:
- Observe two variables as they occur naturally; no manipulation.
- Example: wake-up time and GPA; scatter plot may show a pattern but not causation.
- Limitation: correlation does not imply causation; many extraneous variables (confounds) can explain associations.
- Example extraneous variables: caffeine use, breakfast, sleep quality, mental health, work schedules.
Extraneous/Confounding variables:
- Extraneous variables are unintended factors that could influence results.
- Also called confounds or compound variables.
- Example: caffeine might drive higher alertness, not wake time per se.
Experimental studies:
- Involve manipulation of one variable (independent variable) to observe effects on another (dependent variable).
- Offer a path to causal inferences when well controlled.
- Key features: manipulation, control, random assignment, and careful standardization.
Why experiments help with causality:
- By holding all other factors constant and changing only the IV, differences in the DV can be more confidently attributed to the manipulation.
- However, good experimental design is essential to rule out remaining extraneous variables.

Independent vs. Dependent Variables, and Related Terms

Independent variable (IV): the variable that is deliberately changed/manipulated (predictor).
Dependent variable (DV): the variable that is measured (outcome).
In a simple two-group study: IV defines group membership; DV is the measured score.
Terminology:
- Predictor variable: another term for the independent variable.
- Outcome variable: another term for the dependent variable.
Experimental vs control groups:
- Experimental group: receives the treatment/manipulation.
- Control group: baseline group used for comparison.
Example (video game study):
- IV: type of video game (violent vs. nonviolent).
- DV: number of aggressive behaviors observed.
- Control group: nonviolent game group; Experimental group: violent game group.

Experimental Design and Control of Extraneous Variables

Manipulation: intentionally changing the IV to observe effect on DV.
Control: keeping other variables constant to isolate the IV’s effect.
Possible controls in the video game study:
- Same room, same TV size, same volume, same duration of play, similar age range, and diverse gender/ethnicity in both groups.
Participant variables vs. environmental variables:
- Participant variables: individual differences like age, gender, intelligence, biases.
- Environmental variables: conditions of the experimental setup (time of day, room temperature, equipment quality).
The goal: minimize differences that could confound the interpretation of whether the IV caused changes in the DV.
Ethical concern example: poorly designed or fraudulent studies (e.g., vaccine-autism claim) can cause harm; replication and ethical conduct are critical.

Data Types and Measurement Scales

Discrete vs. Continuous variables:
- Discrete: distinct categories with no intermediate values (e.g., dice roll results 1–6; number of children).
- Continuous: infinite possible values between observed values (e.g., weight, height).
- Rounding/boundaries: in continuous data, scores may be rounded into intervals for ease of reporting (e.g., 149.7 pounds rounded to 150 pounds).
Nominal scales (categorical):
- Categories with names but no intrinsic order or magnitude (e.g., major in college: math, English, psychology).
Ordinal scales:
- Ordered categories with a meaningful order but not necessarily equal intervals (e.g., bronze, silver, gold; small, medium, large).
Interval vs. Ratio scales:
- Interval scales: ordered with equal intervals, but zero is arbitrary (e.g., Fahrenheit temperature; 0°F is not absence of temperature).
- Ratio scales: have a true zero representing absence (e.g., weight in pounds; zero means no weight).
- Important distinction: ratio scales allow meaningful statements about ratios (e.g., twice as heavy), whereas interval scales do not permit such ratio interpretations.
Practical note: understanding scale type determines which statistical methods are appropriate.

Operational Definitions and Measurement Decisions

Variables like hunger or aggression require clear definitions to be measurable.
Example (rats): define hunger as days without food or observed behavior toward food; define aggression as a specific observable behavior (squeaks, bites, etc.).
Transform qualitative concepts into quantitative data to enable analysis.

Practical Takeaways and Real-World Relevance

Why statistics matter in psychology:
- Organize, summarize, and interpret data from experiments, correlational studies, and observations.
- Enable decisions about whether observed differences are likely due to chance or reflect real effects.
- Provide a framework to discuss reliability, validity, and generalizability of findings.
Real-world relevance:
- Describing groups (descriptive) vs. explaining/estimating population parameters (inferential).
- Understanding sampling error helps interpret how confident we should be in study conclusions.
- Ethical considerations and rigorous design prevent misleading claims and promote scientific integrity.

Quick References to Notation and Formulas

Descriptive statistics:
- Mean: $\bar{x} = \frac{1}{n} \sum{i=1}^n xi$
- Range: $R = x{\max} - x{\min}$
- Standard deviation: $s = \sqrt{\frac{1}{n-1} \sum{i=1}^n (xi - \bar{x})^2}$
Z-scores (standardization):
- Population: $z = \frac{X - \mu}{\sigma}$
- Sample: $z = \frac{X - \bar{x}}{s}$
Standard error of the mean: $SE{\bar{X}} = \frac{\sigma}{\sqrt{n}}\quad\text{or}\quad SE{\bar{X}} = \frac{s}{\sqrt{n}}$
Parameter vs. statistic notation:
- Population parameter: $\mu, \sigma^2, …$
- Sample statistic: $\bar{x}, s^2, …$
Hypothesis testing framework (conceptual):
- Null hypothesis: $H_0$
- Alternative hypothesis: $H1$ or $Ha$
- Significance and p-values: probability that data would occur under $H0$ ; small p-values suggest rejecting $H0$ in favor of $H_1$ (thresholds like 0.05 are common, though context matters).
Two-sample comparison (illustrative t-test):
- $t = \frac{\bar{x}1 - \bar{x}2}{sp \sqrt{1/n1 + 1/n_2}}$
- Pooled standard deviation (two-sample case): $sp = \sqrt{\frac{(n1 - 1)s1^2 + (n2 - 1)s2^2}{n1 + n_2 - 2}}$
Confidence intervals (for means; simple form):
- $\bar{x} \pm t^* \frac{s}{\sqrt{n}}$ (where $t^*$ is the appropriate critical value from the t-distribution).

Classroom Notes on Emphasis and Ethics

Expect to learn about normal distribution, z-scores, and distinguishing between descriptive and inferential statistics in upcoming units.
Emphasis on building vocabulary to communicate about data (population vs. sample, parameter vs. statistic, independent vs. dependent variables, extraneous variables, experimental vs. correlational designs).
Ethical considerations: use of data, replication, and avoiding fabrication or manipulation of results; the vaccine-autism example illustrates the consequences of poor design and unethical practices.
The instructor plans to provide additional lecture videos and worked examples to reinforce concepts and help with homework; live attendance remains important for discussion and practice.

Summary of Key Ideas to Remember

Statistics in psychology is a toolkit to organize, summarize, and interpret data from empirical studies.
Descriptive statistics describe data; inferential statistics generalize findings to populations and assess the role of sampling error.
Population vs. sample concepts guide how we estimate population parameters using sample statistics.
Variables can be discrete or continuous and measured on nominal, ordinal, interval, or ratio scales; operational definitions convert abstract concepts into measurable data.
Correlation shows associations but cannot prove causation; experiments use manipulation and control to infer causal relationships, but require careful design to rule out confounds.
Extraneous/compounding variables threaten internal validity; good experimental design seeks to control for these.
Reliability is aided by adequate sample size and representative sampling; ethics and reproducibility are essential to credible science.