KO
Project overview /Pele · SOURCE ATLAS
9 analysis families · 18 conditions

From result to code,
one slide at a time.

Compare each slide with its corresponding Pele map and actual SVG construction branch. Copying the diagram code does not replace statistical computation.

View the 18 original slides ↗ Pele introduction and syntax ↗
CHI

Chi-square

Expected vote distributions differ across party identification
01

Three party groups

Expected vote distributions differ across party identification

944 respondents · ANES96

The association remains visible at the original seven levels
02

Original seven levels

The association remains visible at the original seven levels

944 respondents · ANES96

T-TEST

t tests

Judge the mean-age difference with its interval
03

Independent Welch comparison

Judge the mean-age difference with its interval

944 respondents · ANES96

Paired candidate impressions differ within respondents
04

Paired impression comparison

Paired candidate impressions differ within respondents

944 respondents · ANES96

CORRELATION

Correlation

Age and news use move together without establishing causality
05

Pearson values

Age and news use move together without establishing causality

944 respondents · ANES96

Ranks retain direction and uncertainty
06

Spearman ranks

Ranks retain direction and uncertainty

944 respondents · ANES96

REGRESSION

Regression

Age remains associated with news use after adjustment
07

Classical standard errors

Age remains associated with news use after adjustment

944 respondents · ANES96

Changing error assumptions changes intervals more than estimates
08

HC3 robust errors

Changing error assumptions changes intervals more than estimates

944 respondents · ANES96

ANOVA

One-way ANOVA

Evidence for education-group mean differences is weak
09

Welch + Games–Howell

Evidence for education-group mean differences is weak

944 respondents · ANES96

Report variance assumptions and pairwise corrections together
10

Classical F + Tukey

Report variance assumptions and pairwise corrections together

944 respondents · ANES96

TWO-WAY

Two-way ANOVA

Read nine unbalanced cells with their sample sizes
11

Type III F

Read nine unbalanced cells with their sample sizes

944 respondents · ANES96

Read the joint interaction test before crossing lines
12

HC3 joint F

Read the joint interaction test before crossing lines

944 respondents · ANES96

MODERATION

Moderation

Conditional age associations differ from the interaction test
13

Classical conditional effects

Conditional age associations differ from the interaction test

944 respondents · ANES96

The interaction interval includes zero under robust errors
14

HC3 conditional effects

The interaction interval includes zero under robust errors

944 respondents · ANES96

MEDIATION

Mediation

An indirect model association does not establish causal mediation
15

Monte Carlo intervals

An indirect model association does not establish causal mediation

75 countries · Political Democracy

Indirect intervals depend on method and covariance
16

Delta-method intervals

Indirect intervals depend on method and covariance

75 countries · Political Democracy

SEM

Structural equation models

Separate structural paths, measurement and residuals
17

Six residual covariances included

Separate structural paths, measurement and residuals

75 countries · Political Democracy

Omitting residual covariances changes model fit
18

Residual covariances omitted

Omitting residual covariances changes model fit

75 countries · Political Democracy

Data and validation scope

The examples use an unweighted exploratory analysis of 944 ANES96 respondents and a 75-country Political Democracy model. Expected vote is not actual vote; TV news use is measured in times per week.

The 0.7 skill review draft includes family-specific guidance, reproducible sources and evaluation. The earlier 33 independent numeric checks and browser tests do not replace a new website review. Audience comprehension and physical-device testing remain separate.