Python for Psychological Inquiry

Topics Covered

• What does coding in Python mean
• Introduction to the Python environments: Jupyter Notebooks, Anaconda, and VS Code
• Components of a code editor
• Understanding directories and navigating files

Classroom Activity

• Groups decide their prefered environment and set up on all their devices
• Groups compare layout between enviornments and share pros and cons
• Live demo: setting up a directory in VS Code
• Live demo: starting a new notebook with Google Colab
• Live demo: running Python from the machine terminal

Topics Covered

• Definition and properties of an algorithm: unambiguous, finite, correct
• The gap between human intention and machine instruction
• Why computers need precision that humans take for granted
• Writing and running a first Python statement; understanding the REPL

Classroom Activity

• Unplugged activity: groups write the algorithm for "compute the average Likert score from 10 responses, excluding any blank items"
• Groups compare algorithms; instructor facilitates discussion of where they differ and what each difference implies
• Live demo: show how Python executes each instruction one at a time

R to Python Bridge: Students will recognise the REPL from RStudio's console. R and Python are both interpreted languages — you write a line, the machine runs it. The syntax will look different, but the underlying philosophy is the same.

Topics Covered

• Variables: assignment, naming conventions, reassignment
• Integers and floats: when precision matters (age vs. reaction time in milliseconds)
• Strings: representing names, categories, and text responses
• Booleans: true/false as the basis of all decision-making
• type() and simple type conversion
• Connecting Python types to levels of measurement: nominal (str), ordinal (int), interval/ratio (float)

Classroom Activity

• Students are given a printed dataset card (5 participants, each with age, gender, anxiety_score, above_threshold) and decide which Python type to use for each variable
• Live coding: building the dataset as individual variables, then discussing why this approach becomes unworkable at 500 participants

Discussion: Is a Likert scale score an integer or a float? Does it matter? What about storing someone's grade as 'A' versus as 4? This connects directly to measurement-level concepts from statistics.

Topics Covered

• if, elif, else: structure and indentation
• Comparison operators: ==, !=, <, >, <=, >=
• Logical operators: and, or, not
• Nesting conditions vs. chaining elif
• The concept of a decision boundary and its arbitrariness

Classroom Activity

• Students implement the PHQ-9 scoring rubric as a Python function returning a severity category (minimal, mild, moderate, moderately severe, severe)
• Extension: add validation logic — what should the function return if the score is negative, or greater than 27?

Ethical Touchpoint: Classification systems in psychology have real consequences. A person is not the same as their score, and a cut-off is a convention, not a truth. Code that implements a classification makes the arbitrariness of that convention invisible — a crucial point for future researchers.

Topics Covered

• for loops over lists and ranges
• Accumulators: building a running total or a collection
• while loops: iterating until a condition changes
• break and continue: controlling loop flow
• enumerate() for tracking position
• Nested loops: brief introduction with a concrete example

Classroom Activity

• Given a list of 20 raw test scores, students write a for loop to compute the total and mean without using sum() or len() — building these from scratch with an accumulator
• Extension: loop over the scores again to flag any score more than 1.5 SDs from the mean
• Discussion: why two passes? Could this be done in one pass? What does this tell us about the structure of the computation?

Topics Covered

• def syntax, parameters, and return
• The difference between printing and returning
• Default parameter values
• Docstrings: documenting what a function does, not how
• Calling a function vs. defining it
• Brief preview of the Python Standard Library as 'functions other people wrote'

Classroom Activity

• Students refactor work from Sessions 1.2–1.4 into a small library of functions: compute_mean(scores), compute_sd(scores), classify_severity(score), flag_outliers(scores, threshold)
• Pair exercise: swap code with a partner. Can you understand what each function does from its name and docstring alone?

Unit 1 Closing Reflection: Before this unit, most students experienced programming as typing commands. After this unit, they should experience it as designing a sequence of clearly named, intentional steps.

Topics Covered

• Identifying the problem and breaking it down
• Creating a simple flowchart to address the problem
• Writing pseudo code to mimic the structure of the flow chart

Classroom Activity

• Students are asked to form groups and understand sample problems in data science.
• Guided coding: Each group builds a pseudo code to tackle their problem

Conceptual Bridge: Instead of jumpoing directly into the syntax of Python and the functions at the students' displosal, students are forced to solve problems using logical thinking and reasoning.

Topics Covered

• Lists: ordered, indexed, mutable — for sequences of same-kind data
• List operations: append, remove, slicing, indexing, length
• Dictionaries: key-value pairs — for structured records
• Choosing between lists and dictionaries: a design decision, not a syntax question
• Lists of dictionaries as a natural model for a dataset of participants
• Tuples: briefly — when immutability matters

Classroom Activity

• Students receive a paper participant record and represent it first as a list, then as a dictionary. A second participant is added — discussion follows about which representation scales more naturally
• Guided coding: building a list of participant dictionaries and writing a loop to extract all ages

Conceptual Bridge: In R, the dataframe was given to students from the start. Here, they are building up to it. Understanding why a dataframe is useful requires understanding what it replaces.

Topics Covered

• import pandas as pd: the import statement and the alias convention
• pd.read_csv(): loading real data
• df.head(), df.tail(), df.shape, df.dtypes, df.info()
• Accessing columns and rows: df['column'], df.iloc[], df.loc[]
• Basic filtering: df[df['column'] > value]
• df.describe(): descriptive statistics at a glance — and why glance is not enough

Classroom Activity

• Students load a psychology dataset and answer a checklist of questions before performing any analysis: How many participants? What are the data types? Are there obvious outliers? Are there missing values?
• Discussion: what would happen if we skipped these steps and ran an analysis immediately?

Topics Covered

• What NaN means and how Python/pandas represents absence
• Three mechanisms of missingness: MCAR, MAR, MNAR
• df.isnull(), df.isnull().sum(): quantifying missingness
• Strategies: listwise deletion (dropna), column removal, mean/median imputation
• When each strategy is defensible — and when it is not
• Documenting what you did and why: the research audit trail

Classroom Activity

• Students compare two datasets: one with randomly missing data, one with data missing only for highest-anxiety participants. Could you tell the difference by looking at the NaN pattern alone?
• Apply three different missingness strategies to the same dataset and compare descriptive statistics across the three resulting versions

Ethical Touchpoint: Deciding how to handle missing data is a scientific decision, not a cleaning chore. Both imputation and deletion should be reported transparently, as each carries assumptions about the world.

Topics Covered

• Computing mean, median, and mode from a list using only base Python
• Computing variance and standard deviation step by step: understanding the sum of squared deviations
• sorted() for median; mode as the most common value
• Verifying results with numpy and pandas
• Population vs. sample standard deviation (ddof=0 vs ddof=1)
• Percentiles and quartiles: what they tell us that the mean does not

Classroom Activity

• Paired task: one student writes compute_variance(data, population=True), their partner writes compute_variance(data, population=False). They compare outputs and discuss when each is appropriate — reconnecting to the parameter vs. statistic distinction from the concurrent statistics course
• Visual exploration: scatter plot of raw scores with a horizontal mean line and shaded SD region

Topics Covered

• The matplotlib anatomy: figure, axes, plot elements
• Histogram: choosing bin width and what that choice implies
• Box plot: five-number summary as a visual, interpreting whiskers and outliers
• Bar chart for categorical data: frequency vs. proportion
• Overlaying plots: comparing two distributions visually
• Titles, labels, legends: what every plot must communicate

Classroom Activity

• Students generate three visualisations of the same variable: narrow-bin histogram, wide-bin histogram, and box plot — then write a short reflection on what each emphasises and what it obscures
• Group critique: pairs swap plots and give feedback on whether the visualisation would be clear to a first-time reader

Connection to R: Students have already made plots in R using ggplot2-style thinking. The transition is to understand the underlying figure-and-axes model that both R and Python use. The aesthetic choices are different; the conceptual model is structurally similar.

Topics Covered

• Random number generation: random.random(), random.choice(), random.gauss()
• Simulating coin flips, dice rolls, and then simulated experiments
• The sampling distribution as an empirical object, not just a theoretical one
• Counting how many simulations exceed an observed value: an intuitive p-value
• Why 5,000 simulations give a different answer from 50
• Connecting the simulation result to the formal p-value and alpha concepts

Classroom Activity

• Students simulate: "If sleep quality has no effect on memory test scores (null hypothesis), how often would we see a difference as large as 8 points by chance?" They count the proportion of simulated differences ≥ 8 and compare to alpha = 0.05
• Discussion: what changes if we set alpha to 0.01? What changes if our sample was larger?

Statistical Bridge: This simulation gives students a visceral sense of Type I and Type II errors — a Type I error is what happens when your 'extreme' result is actually one of the 5% of noise outcomes.

Topics Covered

• Revisiting the hypothesis structure: null, alternate, one-tailed, two-tailed
• The logic of the t-statistic: how different are the means, relative to how different we would expect by chance?
• Step-by-step implementation: group means, group variances, pooled standard error, t-statistic
• Degrees of freedom: what they count and why they matter
• Verifying the manual result with scipy.stats.ttest_ind()
• Effect size: why statistical significance is not the same as practical significance

Classroom Activity

• Guided worksheet takes students through each step of the t-test computation; at each step they are asked: what would this number tell us if we stopped here?
• Reflection: the library function returns the same numbers in one line. What do we lose by only ever using the library?

On the Paired t-Test: The paired t-test is introduced as a natural extension — instead of comparing two groups, we compare each participant to themselves across time, reducing to a test of whether the mean difference score is different from zero.

Topics Covered

• Covariance: the numerator of Pearson's r and what it captures
• Why r is bounded between -1 and +1: the normalising role of standard deviations
• Computing Pearson's r step by step, then with numpy.corrcoef()
• Spearman's rho: computing ranks with scipy.stats.rankdata(), then correlating the ranks
• When to use each: normality, outliers, and whether rank or magnitude is meaningful
• The correlation matrix: representing many relationships at once

Classroom Activity

• Students compute both Pearson's r and Spearman's rho for the same two variables in a dataset containing strong outliers, then discuss: which better represents the true relationship?
• Scatter plot with correlation coefficient annotated: students produce this for three pairs of variables and rank them by strength, then discuss whether strong correlation implies importance

Topics Covered

• Classic spurious correlations and the third-variable problem
• The concept of a confounding variable: what is the third thing?
• Partial correlation: holding a third variable constant — conceptual introduction
• Directionality: when A correlates with B, could B cause A?
• The ecological fallacy: correlations at the group level may not apply to individuals
• Implications for psychological research: media misinterpretation of correlation findings

Classroom Activity

• Students identify a plausible confound in a dataset with a strong correlation, subset the data by that confound, and compute the correlation separately within each subset
• Writing exercise: write a single paragraph a journalist might write about the finding, then a corrected version written by a researcher

Topics Covered

• From correlation to regression: the line of best fit as a predictive model
• The slope and intercept: what each means in the context of the variables
• Residuals: the difference between prediction and reality
• R-squared: what proportion of variance in Y is explained by X?
• Chi-square: testing whether the distribution of one categorical variable is independent of another
• Observed vs. expected frequencies; computing chi-square manually in Python

Classroom Activity

• Regression: students implement the OLS formula for slope and intercept manually, plot the resulting line over a scatter plot, then use sklearn to verify
• Chi-square: given a contingency table (gender × diagnosis), students compute expected frequencies and chi-square without scipy, then verify with it

Concurrent Statistics Link: Both regression and chi-square are being introduced simultaneously in the statistics course. This session gives students a concrete computational model that will deepen their understanding of the formal treatment they are receiving in parallel.

Topics Covered

• Code comments: when to comment and what not to comment
• Docstrings: documenting functions for others
• Variable naming as documentation: why score_before_training > s1
• The analysis pipeline as a sequence: import, inspect, clean, describe, analyse, visualise, interpret
• Jupyter Notebooks as a format for literate programming
• Setting a random seed for reproducibility in simulations
• The replication crisis in psychology: a brief contextual discussion

Classroom Activity

• Students receive a deliberately poorly documented script that performs a correct analysis. They reconstruct what it does, add appropriate comments and docstrings, rename variables, and reorganise it — without changing any analytical logic
• Pair exercise: one student runs the partner's refactored script and records every point of confusion. These confusions become the agenda for a short revision

Topics Covered

• Truncated y-axes: making small differences look large
• Cherry-picking time windows in a time-series plot
• Inappropriate chart types: 3D pie charts, dual-axis plots with arbitrary scales
• Missing error bars or confidence intervals
• Colour choices that mislead: gradients implying order where there is none
• APA and ethical reporting standards for figures

Classroom Activity

• 'Spot the Manipulation': students are shown six published-style charts containing at least one misleading choice each, and must identify the problem and suggest a correction
• Fix it: students recreate two of these charts correctly in Python, writing a one-sentence APA-style caption for each

A misleading visualisation is not always the result of bad intent. Often it reflects a failure to think from the reader's perspective, or an uncritical acceptance of software defaults.

Topics Covered

• What a results section should contain: statistic, df, p-value, effect size, direction
• Translating statistical language for non-specialist audiences
• Common misinterpretations: what a p-value is not, what 'no significant difference' does not mean
• Reporting limitations: what the analysis cannot tell us
• Connecting to ethical data reporting standards from the R programming course

Classroom Activity

• Given output of a correlation analysis (r = 0.43, p = 0.02, n = 36), students write: (a) a formal APA-style results paragraph, and (b) a three-sentence plain-language explanation for a school newsletter
• Pairs read each other's versions and discuss where the technical version is inaccessible and where the plain-language version has been oversimplified

Pipeline Steps

1. Load and inspect the dataset: shape, types, missing values
2. Clean: address issues found in step 1; document all decisions
3. Describe: compute and display relevant descriptive statistics
4. Visualise the distributions of key variables
5. Formulate and test a hypothesis appropriate to the research question
6. Visualise the relationship of interest
7. Interpret: what does the analysis suggest? What can it not tell us?
8. Reflect: what would you do differently with more time or more data?

Instructor Role: The instructor does not answer "how do I code X?" questions in this session. Instead, questions are redirected: "What are you trying to find out at this step? What would the answer look like?" This scaffolds independent problem-solving.

Structure of the Section

• Brief presentations: each student or pair shares one visualisation and one finding from Section 4.4 (20 minutes total)
• Structured written reflection responding to the prompts below
• Whole-group discussion: what concepts from statistics do you now understand differently?
• Looking ahead: brief introduction to what Python can do that this course has not covered — NLP, web scraping, machine learning — as motivation for continued learning

Closing Note: This course has not made you a programmer. It has done something more specific: it has given you a way of taking a problem apart, representing it precisely, and testing your understanding of it by trying to make a machine do it. The tools will change. The way of thinking does not have to.