Course · Intermediate · Statistics and Modelling
Applied Statistics for Programmes
Uncertainty, comparison and significance, taught around the claims a programme report makes — not around a textbook sequence of tests.
What you will be able to do
- Read the distribution of an indicator before computing anything on it, and say what its mean is hiding
- Put a confidence interval on a proportion and write the result as an interval in a report sentence
- Compare two groups with the right test, state its assumptions, and check them against the data
- Separate statistical significance from operational relevance, and report an effect size beside every p-value
- Recognise when the unit of analysis is not the row, and correct the sample size accordingly
- Run a family of comparisons without manufacturing the false positives it invites
Standards and methodologies
Module 4 spent six courses deferring to this one. Every time a lesson said “put an interval on that before writing the sentence”, it was pointing here — and the gaps it deferred are the worked examples, already computed on files you have already cleaned.
This is not a statistics textbook with programme data as decoration. It is organised around the four claims a programme report actually makes: this is the level, this group differs from that one, this changed, and this caused that. Each unit takes one and works out what it costs to say it honestly.
Three results shape the course, all computed from committed files.
The mean E. coli count in tested households is 29.6 CFU/100 mL and the median is 0. More than half of tested households have no detectable contamination and the mean is describing a tail that reaches 608. A report quoting the mean has described no household in the sample.
Two gaps from module 4, one real and one not. Boys are over-age at 39.2% against girls at 33.8% — a 5.5-point gap whose interval runs from −0.4 to +11.3, so the honest answer is that this survey cannot tell. Cases reporting a disability complete a referral at 26.7% against 46.2% — a 19.4-point gap with an interval from −26.1 to −12.8, which no reasonable reading makes zero.
1,200 students are 24 schools. Not one of the twenty-four schools has both fed and unfed students, so a comparison treating children as independent observations has an effective sample size of 24 and a standard error 1.8 times too small.
Both Python and R throughout. You should have done modules 3 and 4 — this course assumes you can defend a denominator, and it spends its time on what to do once you have one.
Practice
Reading the lessons is not the same as having done the work. Each of these applies the course to a dataset it did not teach on.
- Lab · 150 minThe district with no open defecationA district table shows six districts, one of them with zero open defecation and a significant difference against its neighbour. There are three districts, the difference is a spelling variant, and the interval on the zero reaches 16%.
- Exercise · 60 minNine points on a different testA funding proposal reports a 9.7-point literacy gain at p < 0.001. The baseline paper had 40 items and the endline paper had 50, 156 of the weakest students never sat the second one, and the test treated the same children as two independent samples.
Progress
Take it offline
The whole course as a typeset PDF — every lesson, every code example, the data dictionary and the indicator definitions. Generated from the same source as this page.
The LaTeX source ships alongside each PDF, so an organisation can rebrand the handout or fold a lesson into its own training pack.