Descriptive statistics: describe the data.
Summarize (statistics), tabulate (frequency distribution), graph (histogram, etc.)

Data

Levels of measurement

Level		Examples	What can do with
Nominal	names, labels, categories	Yes/No, Agree/Disagree, Have/Havenot, Success/Failure, M/F, ... MaritalStatus, State, County, Zipcode, Major, Brand,make,model,color, Place race,religion,party,ideology..., TaxFilingStatus, Blood type, Housing type, Pet	Count/tally each category. Relative frequency. Mode. Bar chart. Chi-square Tests (independence, goodness-of-fit) Confidence interval 1-PropZInt
Ordinal	orderable/rankable categories but differences (obtained by subtraction) between data values either cannot be determined or are meaningless.	class(frosh/soph/jun/sen), trim levels, film ratings, gold/silver/bronze, letter grades, days of week, months, Education level, clothing sizes, pain scales, military rank, star ratings, priority/risk levels Percentiles. Likert scale: Strongly disagree / Disagree / Neutral / Agree / Strongly agree Very dissatisfied / Dissatisfied / Neutral / Satisfied / Very satisfied Poor / Fair / Good / Very good / Excellent	Above + median/quartiles, Spearman.
Interval	Numbers: orderable, and differences between data values can be found and are meaningful. But no natural zero (meaning none of the quantity).	Temperature C or F, Years/Dates, shoe size, IQ/SAT, FICO, pH 0 is fakish	histogram, mean, median, SD... Estimation, CI: t-test, Hypothesis testing, ANOVA, correlation, linear regression
Ratio	Numbers: orderable, and differences between data values can be found and are meaningful, and natural zero (meaning none of the quantity), and ratios (eg. "twice as much") are meanginful.	Weight Height Age Length Area Volume Time Money TemperatureK Energy BP LDL BMI DJI S&P500	Above + CV, GM,

Example: Population: weights of adults in country/county.
Not possible to census this. So need a non-biased, representative sample (a teaspoon of the pot of soup).
Ideal: Simple random sample (SRS): every adult equally-likely to be in the sample and every sample of that size is equally-likely.
  Bad: voluntary response, convenience sample.
Collect data. Measured vs self-reported (unreliable).
Calculate/derive statistic from the data: a point estimate of the parameter. But samples have uncertainty/variability so determine [confidence] interval estimate.
Inferential statistics: use probability to understand/quantify/describe uncertainty.
If have census, i.e. population is all known, no need to sample, just describe the population. Sample(s) only useful/taken/needed to estimate population parameter(s).

Data set mapped/transformed/normalized to z-scores: each datum x: its number of SDs from the mean:
z = (x-mean) / SD
Within ±2 is "normal". [2,2]
≤-2 (-∞,-2] or ≥2 [2,∞) is "statistically significant", i.e. maybe important.