freqweights (version 1.0.1)

lmfreq: lmfreq is used to fit linear models with frequency tables

Description

To fit linear models with data grouped in frequency tables.

Usage

lmfreq(formula, data, freq = NULL)
.lmfreq(formula, tfq)
## S3 method for class 'lmfreq':
logLik(object, ...)
## S3 method for class 'lmfreq':
extractAIC(fit, scale = 0, k = 2, ...)
## S3 method for class 'lmfreq':
AIC(object, ..., k = 2)
## S3 method for class 'lmfreq':
nobs(object, ...)
## S3 method for class 'lmfreq':
summary(object, ...)
## S3 method for class 'lmfreq':
print(x, ...)
## S3 method for class 'summary.lmfreq':
print(x, digits = getOption("digits") - 3, ...)
## S3 method for class 'lmfreq':
predict(object, ...)

Arguments

formula

an object of class formula

data

a data frame that must contain all variables present in formula and freq

freq

a character string specifying the variable of frequency weights

tfq

a tablefreq object

object

a lmfreq object

...

See Details

fit

a lmfreq object

scale

not used

k

penalty parameter

x

a lmfreq object

digits

digits

Value

• It returns an object of class lmfreq, very similar to lm

Details

It computes the linear model of a frequency table. See lm for further details.

Any variables in the formula are removed from the data set.

The dot function are for programming purpose. It does not check the data.

See Also

tablefreq

Examples

## Benchmark
if(require(hflights)){
  formula <-  ArrDelay ~ DepDelay
  print(system.time(a <- lm(formula, data=hflights)))  ## ~0.4 seconds
  print(system.time(b <- lmfreq(formula, data=hflights))) ## ~0.12 seconds. 4x faster
}

l0 <- lm(Sepal.Length ~ Sepal.Width,iris)
summary(l0)

tfq <- tablefreq(iris[,1:2])
lf <- lmfreq(Sepal.Length ~ Sepal.Width,tfq, freq="freq")
summary(lf)

all.equal(coef(lf),coef(l0))
all.equal(AIC(lf),AIC(l0))

newdata <- data.frame(Sepal.Width=c(1,NA,7))
predict(lf, newdata)

if(require(MASS)){
   stepAIC(lf)
}

system.time(lmfreq(Sepal.Length ~ Sepal.Width,tfq, freq="freq"))
system.time(.lmfreq(Sepal.Length ~ Sepal.Width,tfq)) # Fast

library(dplyr)
igrouped <- iris %>% group_by(Species)
models <- igrouped %>% do(model=lmfreq(Sepal.Length ~ Sepal.Width, .))
coefs <- models %>%
  do(cbind(as.data.frame(rbind(coef(.$model))),
           Species=.$Species))
coefs

## If data is too granular, benchmark is worst
n <- 10^6
data <- data.frame(y=rnorm(n),x=rnorm(n))
system.time(lm(y~x,data)) ## ~5 seconds
system.time(lmfreq(y~x,data)) ## ~ 15 seconds
system.time(tfq <- tablefreq(data)) ## ~ 5 seconds
nrow(tfq) # same number of rows than original data
system.time(.lmfreq(y~x,tfq)) ## ~ 10 seconds