Experimental Design in Python
| Course Details | ||||||||
|---|---|---|---|---|---|---|---|---|
|
||||||||
| Languages | ||||||||
|
||||||||
| Outcomes | ||||||||
|
||||||||
| Price & Terms | ||||||||
|
||||||||
| Technical | ||||||||
|
||||||||
| Specification | ||||||||
|
||||||||
Learn to set up blocked and factorial experiments in Python, choose the right statistical test, and size samples with power analysis.
Experimental design decides whether a test result can be trusted: how subjects are assigned to groups, which statistical test fits the data, and how many observations are needed to detect a real difference. You work in Python on data such as marketing click-through rates, crop yields and loan approvals. Along the way you build blocked and factorial experiments, compare groups, size samples with power analysis, and deal with data that breaks the usual test assumptions.
What You Will Achieve
- Arrange an experiment so that treatment groups are comparable, and confirm with plots and tests whether the data look normal.
- Match a t-test, ANOVA or chi-square test to the question and data type, then pinpoint which pairs of groups differ after an ANOVA.
- Measure how far groups differ with Cohen’s d and work out the smallest sample that reaches a target statistical power.
- Detect unequal variance and interactions between factors, and choose a suitable nonparametric test when parametric assumptions fail.
Program Overview
Setting Up Experiments and Design Techniques
Lays the groundwork for a sound experiment, then moves to designs that control variance and adjust for covariates.
- Random and non-random assignment of subjects
- Blocking and stratifying experimental data
- Visual and analytical normality checks
- Factorial designs, randomized block designs and covariate adjustment
Statistical Tests and Power Analysis
Selects and runs the tests that compare groups, follows up significant results, and plans sample sizes.
- Choosing the right test for petrochemical, human resources and finance cases
- Post-hoc analysis with Tukey’s HSD after ANOVA
- P-values, alpha, significance levels and errors
- Power analysis for sample and effect size
Complexities in Experimental Data
Handles messy experimental results and turns them into findings presented to different stakeholders.
- Synthesizing insights with pandas and communicating them to stakeholders
- Checking heteroscedasticity and transforming data
- Mann-Whitney U testing of food preservation methods
Certificate Awarded (Paid)
Job Outlook
These skills apply wherever groups are compared to judge a change, including marketing and product analytics, agricultural trials, lending decisions, human resources studies, and chemical or food quality testing.
Good for: Analysts and researchers who run A/B tests or trials and need to choose a design, a test and a sample size they can defend.
Keep in mind: Tests run in SciPy with some statsmodels functions on supplied datasets, so collecting data for an experiment of your own is left to you.
Bottom line: A practical run through experimental design choices in Python that adds power analysis and nonparametric options to the usual t-tests and ANOVA.
Frequently Asked Questions
Which Python libraries do the exercises use?
Most statistical tests are run with SciPy, including scipy.stats for checks such as heteroscedasticity, while several functions come from statsmodels. Pandas is used to combine and summarize results from complex experiments.
What kinds of datasets appear in the exercises?
The practice data spans many fields: marketing click-through rates, crop yields, loan approval rates, athletic performance, chemical reaction rates, employee productivity, chick growth, toy durability and the shelf life of preserved food.
How does the course handle data that fails test assumptions?
Normality is checked both visually and with analytical tests, and unequal variance is tested before analysis. When the data still do not suit a parametric test, the lessons transform it or apply a nonparametric test such as Mann-Whitney U.
User Reviews
Write your review
You must be logged in to post a review.
Reviews can only be posted by registered learners. Log in or create a free account, then you will return here to post.
There are no reviews yet.