The Basic Research Workflow
Core summary
The research workflow proceeds from question to design to data collection to analysis to interpretation to publication. The analysis plan should be determined by the research design, not chosen after seeing the data.
Detailed explanation
Detailed explanation
This lesson connects the dots between what you have learned so far and the statistical analysis that comes later in Level 6. The workflow is: (1) Define your research question (PICO). (2) Formulate hypotheses (H₀ and H₁). (3) Choose a study design (Level 4). (4) Determine your primary outcome and data types. (5) Select the appropriate statistical test BEFORE collecting data. (6) Calculate sample size based on your expected effect. (7) Collect data. (8) Clean and prepare data. (9) Run your pre-planned analysis. (10) Interpret results (effect size, CI, p-value). (11) Write and publish. The critical point is that the analysis plan is determined by the design, not the other way around. You choose your statistical test based on your research question, data types, and study design — before you collect a single data point. This prevents the temptation to try multiple tests until one gives a significant result (p-hacking). The flowchart for choosing a statistical test follows a simple logic: What is your research question? (comparing groups, testing association, predicting outcome). What type is your outcome variable? (continuous, categorical, binary, time-to-event). How many groups are you comparing? (two, more than two). Are your data paired or independent? These questions lead to a specific test — which you will learn in Level 6. For now, understand the principle: design drives analysis. Not the reverse.
Clinical example
A surgeon plans to compare operative times between two surgical techniques. She determines: question = 'Is technique A faster?', design = prospective cohort, outcome = operative time (continuous), groups = 2 (independent). Before collecting any data, she pre-specifies an independent samples t-test (or Mann-Whitney if data are non-normal). She calculates that she needs 40 patients per group. Only then does she begin.
Research example
Pre-analysis plans (PAPs) are increasingly required by journals and registries. A PAP specifies exactly which analyses will be run before data collection begins, preventing post-hoc manipulation of results.
Knowledge check
Q1. When should you choose your statistical test?
Q2. It is acceptable to try multiple statistical tests until you find a significant result.
Q3. What determines which statistical test to use?