Volume 8 · Designs, and the domains with their own statistics

The analysis cannot rescue a design. It can only describe one.

Twelve days on matching the statistics to the study: what licenses a causal claim, the four validities, counterbalancing, quasi-experiments and what propensity scores really do, single-case experiments, longitudinal change and attrition, time series, survey missingness, reaction times, signal detection, and the statistics of imaging and EEG.

12
days of 45–55 min
4
kinds of validity
7
threats to internal validity
3
missingness mechanisms
{{ d.badge }} {{ d.meta }}

{{ d.title }}

{{ d.lead }}

{{ s.t }}
{{ b.title }}
{{ s.t }}
The mistake this topic produces
{{ s.t }}
{{ c.eyebrow }}
{{ c.title }}
{{ s.t }}
{{ h.label }}
{{ c.v }}
{{ b.eq }} {{ b.plain }} {{ sy.d }} {{ b.work }}
Do this now · 15 minutes
{{ d.task }}
Reference

{{ rf.title }}

{{ rf.lead }}

{{ h.label }}
{{ c.v }}
Checkpoint

Ten questions before you move on

Design questions carry as many marks as computation in most entrance papers, and they are the questions a viva actually turns on.

{{ quizCounter }}
{{ quizScore }}
{{ quizQ }}
{{ quizFb }}

Next: volume 9, Modern methods

Twelve days on what doctoral work runs on now: mixed-effects models for nested and trial-level data, Bayesian inference and Bayes factors, MCMC, meta-analysis and publication bias, robust and resampling methods, simulation as a design tool, and causal inference with DAGs.

Start day 101 →