# glm does generalised linear models in R
# some fake data
x1 <- 1:50
x2 <- rep(1:5, 10)
y <- rnorm(50, mean = 1/(2 * x1 + 4 * x2), sd = 0.05)
df <- data.frame(y = y, x1 = x1, x2 = x2)
# formula species the linear predictor - same syntax as lm

# family specifies the distribution of the response and the link function
# they are some common defaults for usual dists
# look at 
?family
# for more details

# i.e. normal linear regression
glm(y ~ x1 + x2, data = df, family = gaussian)

# normal linear regression with inverse link
glm(y ~ x1 + x2, data = df, family = gaussian(link = "inverse"))
fit <- glm(y ~ x1 + x2, data = df, family = gaussian(link = "inverse"))
anova(fit)

# residuals for a glm fit can be of different types
# type = c("deviance", "pearson", "working", "response", "partial"), ...)

# deviance residuals
residuals(fit, type = "deviance")