Population, Samples, Variables & Data

Slide 1

Population, Samples, Variables & Data

Slide 2

Where we left it

Slide 3

What you will be able to do at the end of today

Slide 4

Act I · Your Question, Turned Into Rows and Columns

Slide 5

Your question

Slide 6

Every study becomes a rectangle

Before any test, the data have a shape. Rows are the units you measured. Columns are the things you measured on them. The rest of the session lives inside that picture.

Slide 7

Two questions to ask before any other

Slide 8

So what is one row?

A row is not automatically a person. It is the unit the researcher decided to measure. Eyes, visits, households, and patients are different rows, and mixing them is how a paper quietly changes what it is about.

Slide 9

The row you measured is not always the row you treated

Slide 10

And what is one column?

A column is one measured thing, with a type and a job. Until those two are named, a later table is just a pile of numbers with headings.

Slide 11

Act II · Who the Study Is Actually About

Slide 12

Why not simply measure everybody?

Slide 13

Four populations, not one

Target, source, sampled, and analyzed are not synonyms. Papers collapse them into one word. The session pulls them apart so a later claim about ‘the population’ has to say which one.

Slide 14

Who was let in, and who was kept out

Slide 15

How those 2,039 were picked

Slide 16

Who was never on the list at all

Slide 17

The 229 people who never arrived

Slide 18

Parameter and statistic

The parameter lives in the population you did not measure. The statistic lives in the sample you did. Session 1’s backwards arrow now has names on both ends.

Slide 19

Two worlds, and one risky arrow

Everything later (standard error, confidence intervals, p values) is an attempt to say how much risk that arrow carries. Those tools are named and left. Today only names the two worlds.

Slide 20

The notation, once, so it stops being decoration

Slide 21

Three samples, three answers

Slide 22

One substitution, so the symbols have somewhere to land

Slide 23

Act III · What Kind of Thing Is This Column, and What Job Does It Do?

Slide 24

Back to the rectangle, because the rest of today lives in it

Slide 25

What kind of thing is this column?

Slide 26

The boundary that is genuinely arguable: ordinal or discrete?

Slide 27

Sort these eight

Slide 28

The answers, and the three that are worth arguing about

Slide 29

The same variable, twice, in two different types

Slide 30

What a threshold costs you

Slide 31

What the cut bought

Slide 32

Where a column comes from

Slide 33

Four jobs a column can do

Outcome, exposure or predictor, confounder, and descriptor. The same blood-pressure column can do more than one job in two different papers. The job is a decision, not a property of the variable.

Slide 34

Exposure is a causal word. Predictor is not.

Calling a column an exposure is already a claim about the world. Calling it a predictor is not. The session makes that distinction early so later regression language cannot hide it.

Slide 35

Act IV · The Role Is Not in the Column. It Is in the Question.

Slide 36

The same columns, two questions

Slide 37

Association, and what it is not

Slide 38

Four reasons two columns move together

Slide 39

In this study, smoking looks protective

Among men in the Sylhet survey, current smokers had lower odds of hypertension. Nobody in the room believes smoking prevents high blood pressure. The point is the habit: an association is not a cause, and a paper that looks surprising is where the four reasons two columns move together become useful.

Slide 40

Close

Slide 41

This is what you are going to write

Slide 42

The variable table has a professional name

Slide 43

Before the next session