Appendix C — Glossary and R reference

C.1 Statistical language

Term	Meaning in this book
Population	The set of units about which a question is asked; define its crop, place, time, and management scope.
Sample	Observed units selected from a population. Convenience samples can limit generalization.
Experimental unit	Smallest unit independently assigned a treatment.
Observational unit	Unit on which a measurement is recorded; it may be a subsample of an experimental unit.
Replication	Independent repetition of treatment assignment at the experimental-unit level.
Randomization	A chance-based assignment procedure that helps separate treatment from nuisance influences.
Blocking	Grouping comparable experimental units before randomization to account for known variation.
Confounding	A situation where influences vary together so their separate contributions cannot be identified from the comparison.
Parameter	A quantity describing a population or a model, such as a mean or regression slope.
Estimate	A value calculated from data to represent a parameter.
Residual	Observed response minus its fitted model value.
Standard deviation	Spread among observations, in the measurement’s units.
Standard error	Estimated sampling variation of a statistic, such as an estimated mean.
Confidence interval	An interval constructed by a procedure with stated repeated-sampling coverage under assumptions.
Prediction interval	An interval for a new response, including individual-response variation under a model.
Effect size	Magnitude of a difference or association; its practical relevance depends on context and units.
p-value	Tail probability for a statistic under a specified null model and assumptions.
Interpolation	Prediction within the observed predictor domain.
Extrapolation	Prediction beyond the observed domain; it requires additional justification.
Holdout	Observations reserved from fitting for assessing prediction.
Reproducibility	Ability to regenerate an analysis from its inputs, code, and environment.

C.2 R language reminders

Task	Expression	Caution
Assign	yield <- 4.8	Names are case sensitive.
Compare equality	x == 5	= also occurs in function arguments; it is not an equality test.
Combine values	c(1, 2, 3)	Atomic vectors coerce to a common type.
Make a sequence	seq(0, 10, by = 2)	Choose spacing appropriate to units.
Select vector entries	x[c(1, 3)]	Indices begin at 1.
Select a table	dat[rows, columns]	The comma distinguishes rows from columns.
Extract a column	dat$yield_t_ha	Explicit names make provenance clearer.
Detect missingness	is.na(x)	Do not compare with x == NA.
Inspect	str(dat), summary(dat)	An import needs validation beyond these summaries.
Group and summarize	group_by(), summarise()	Summary rows represent a new observational level.
Fit a linear model	lm(y ~ x, data = dat)	Formula terms express a model, not causal proof.
Fit ANOVA	aov(y ~ treatment, data = dat)	The formula must represent the design.
Fit nonlinear growth	nls(y ~ formula, data = dat, start = ...)	Convergence alone is insufficient.
Inspect help	?function_name	Read arguments, returned values, assumptions, and examples.

C.3 Common errors and a repair sequence

When a command fails, read the first error, check spelling and object existence, inspect types and dimensions, and run the smallest relevant part. For file errors, check getwd() and file.exists(). For a missing package, check installation and loading. For a model failure, inspect the data and the model assumptions before changing optimizer settings. Keep a short note of the cause and resolution; debugging is a skill developed through explanation.

C.4 Before you submit an analysis

State the question and unit; document the source and units; retain raw data; record cleaning decisions; show distributions; justify the model; examine residuals; report effects and uncertainty; distinguish association from causation; and render from a fresh session. These steps are scientific habits, not a substitute for judgment.