R/testAR-package.r

In adrowe1/test: how to make a package

#' testAR
#'
#' @name testAR
#' @docType package

library( "ggplot2" )

qqplot.data <- function (vec) 
{
  y <- quantile(vec[!is.na(vec)], c(0.25, 0.75))
  x <- qnorm(c(0.25, 0.75))
  slope <- diff(y)/diff(x)
  int <- y[1L] - slope * x[1L]

  d <- data.frame(resids = vec)

  ggplot(d, aes(sample = resids)) + stat_qq() + geom_abline(slope = slope, intercept = int)

}

#does this commit?

compare.models <- function ( raw ) {
      dat <- reshape2:::melt.data.frame(raw, id.vars=c("type", "time"), variable.name="measurement", value.name="uptake")

    #ggplot(dat, aes(x=time*2, y=uptake, colour=type) ) + geom_point() + geom_smooth(method="lm")

    model.minimal <- lm(uptake ~time, dat)
    model <- lm(uptake ~time * type, dat)

    #summary(model)

    # qqplot.data(model$res)

    anova(model.minimal, model)
}

                                    



#dat$pred <- predict(model, dat)
#dat$residuals <- dat$uptake-dat$pred

 
plotit <- function ( raw ) {
        dat <- reshape2:::melt.data.frame(raw, id.vars=c("type", "time"), variable.name="measurement", value.name="uptake")              
                                                
        model <- lm(uptake ~time * type, dat) 
        dat$pred <- predict(model, dat)
        dat$residuals <- dat$uptake-dat$pred
        
        # ggplot(dat, aes(x=residuals, fill=type)) + geom_histogram(binwidth=3, position="dodge")
        ggplot(dat, aes(x=time*2, y=residuals, colour=type) ) + geom_point() + geom_smooth()
        
        
}



# dat$res.class <- "mid"
# dat$res.class[dat$residuals>5] <- "high"
# dat$res.class[dat$residuals<(-3)]<- "low"
# 
# 
# model2 <- lm(uptake ~ time * type, dat[dat$time>4,])
# model3 <- lm(uptake ~ time, dat[dat$time>4,])
# anova(model2, model3)
# 
# qqnorm(model2$res)
# qqnorm(model3$res)