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

In freqweights: Working 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

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.

Value

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

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

## Not run: 
## 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

## End(Not run)

freqweights documentation built on May 29, 2017, 12:01 p.m.