1 Codebook — Group 5 — Heart failure + Asia map (India-facing stand-in)

Spine: Who dies after heart failure, who grew up poor in the GSS, and where in Asia is life short?

1.0.1 Panel A

cardiac_failure recoded 0/1 to yes/no; Quants: age, ejection_fraction, serum_*, platelets, time

n = 299 rows, 13 columns.

Column Inferred type Notes
age Quant 47 unique, 0 NA
anaemia Qual 2 unique, 0 NA
creatinine_phosphokinase Quant 208 unique, 0 NA
diabetes Qual 2 unique, 0 NA
ejection_fraction Quant 17 unique, 0 NA
high_blood_pressure Qual 2 unique, 0 NA
platelets Quant 176 unique, 0 NA
serum_creatinine Quant 40 unique, 0 NA
serum_sodium Quant 27 unique, 0 NA
sex Qual 2 unique, 0 NA
smoking Qual 2 unique, 0 NA
time Quant 148 unique, 0 NA
DEATH_EVENT Qual 2 unique, 0 NA

1.0.2 Panel B

AER GSS7402 Quals + education-as-years (Quant). Do not average ethnicity.

n = 9,120 rows, 6 columns.

Column Inferred type Notes
kids Quant 9 unique, 0 NA
education Quant 21 unique, 0 NA
ethnicity Qual 2 unique, 0 NA
city16 Qual 2 unique, 0 NA
lowincome16 Qual 2 unique, 0 NA
immigrant Qual 2 unique, 0 NA

1.0.3 Panel C

gapminder 2007, Asia only — closest frozen stand-in for an NFHS/malaria map. Join to rnaturalearth. India is one row.

n = 33 rows, 6 columns.

Column Inferred type Notes
country Qual 33 unique, 0 NA
continent Qual 1 unique, 0 NA
year Quant-looking — check levels; may be Qual 1 unique, 0 NA
lifeExp Quant 33 unique, 0 NA
pop Quant 33 unique, 0 NA
gdpPercap Quant 33 unique, 0 NA

1.1 Pair card (Day 1 brick)

Pair Legal question First mark Later test (kit first, campus later)
Quant + Qual (2 levels) Does amount differ by kind? Two violins Two-sample comparison
Quant + Qual (3+ levels) Do kinds have different amounts? Grouped violin ANOVA
Qual + Qual Are kinds entangled? Mosaic Chi-square / two proportions
One Qual vs “more than half” Is this share rare? Bar / waffle One proportion
Quant + Quant How do two amounts move? Scatter Not a Day-7 campus hunch unless they insist

1.2 Forbidden list

  • Do not histogram a yes/no, a 0/1, a chest-pain type, a grade, or a wealth-like code.
  • Do not average Sex, ChestPain, degree, helmet, immigrant, ethnicity.
  • Do not choropleth a count when a rate exists (infant mortality, under-5 mortality, lifeExp are rates or expectations — pop is a count).
  • Do not run ANOVA on Panel B.
  • Do not treat area in the Scottish file as a map.
  • One AI prompt per panel, after paper, and only for a geom the codebook allows.

1.3 Sources (cite on the Quarto site)

  • CardioDataSets (Cáceres Rossi): heartdisease_tbl_df, cardioRiskFactors_df, scottish_CVD_df, cardiac_failure_df
  • openintro: infmortrate, sowc_child_mortality, bdims, yrbss, gss2010
  • gapminder: gapminder
  • AER: GSS7402
  • Files pulled via Rdatasets (Arel-Bundock)

NHANES and NFHS-5 / malariaAtlas were the preferred A and India-C objects. They are not in this zip because they need extra packages or an API. Group 3 A is a body-measure stand-in. Group 5 C is Asia-only gapminder, not district NFHS.

Back to top