Explore
Find agent skills by outcome
123,538 skills indexed with the new KISS metadata standard.
- Default method is Nelder-Mead (no gradients
robust but slow). Poor for 1D — use Brent or optimize().
- predict.glm with type = response gives predictions on the response scale (e.g.
probabilities for logistic); type = link (default) gives on the link scale.
- lowess parameter is f (span
default 2/3); loess parameter is span (default 0.75).
- lowess is the older function; returns `list(x
y)` — cannot predict at new points.
- Direction: direction = both (default)
forward
- Self-starting models (SSlogis
SSasymp
- aov is a wrapper around lm that stores extra info for balanced ANOVA. For unbalanced designs
Type I SS (sequential) are computed — order of terms matters.
- For Type III SS
use car::Anova() or set contrasts to contr.sum/contr.helmert.
- Quasi-families (quasibinomial
quasipoisson) allow overdispersion — no AIC is computed.
- binomial accepts response as: 0/1 vector
logical
- `anova(glm_obj
test = Chisq) for deviance-based tests; F` is invalid for non-Gaussian families.
- weights in glm means prior weights (not frequency weights) — for frequency weights
use the cbind trick or offset.
- Column names from model.matrix can be surprising: e.g.
factorLevelName concatenation.
- Common families: binomial(link = logit)
poisson(link = log)
- Default family = gaussian(link = identity) — glm() with no family silently fits OLS (same as lm
but slower and with deviance-based output).
- I() is required to use arithmetic operators literally: y ~ x + I(x^2). Without I()
^ means interaction crossing.
- terms object attributes: order (interaction order per term)
intercept
- model.matrix creates the design matrix including dummy coding. Default contrasts: contr.treatment...
contr.poly for ordered.
- Captures output from cat
Non-obvious behaviors
gotchas
- Converts character vectors to appropriate types (logical
integer
- as.is = TRUE (recommended): keeps characters as character
not factor.
- `formatC(x
format = f