Statistic summary in R

• summary the statistics of data

• visualize the statistics (boxplot and histogram)

1. view the data

   library("AzureML")
   ws <- workspace()
   dat <- download.datasets(ws, "Automobile price data (Raw)")
   head(dat)

   results:

2. process the dataframe:

   cols = c('price','bore','stroke','horsepower','peak.rpm')
   ## convert ? to an NA
   dat[,cols] = lapply(dat[,cols], function(x) ifelse(x == '?', NA, x))
   ## remove rows with NAs
   dat = dat[complete.cases(dat),]
   ## Covert character columns to numeric
   dat[,cols] = lapply(dat[,cols], as.numeric)
   str(dat)

   complete.cases(dataframe) keep those rows without NA. lapply(dataframe, function) is used to manipulate the dataframe globally. The statement of the processed dataframe: 

3. view the summary of the dataframe: 

   describe = function(df, col){
   tmp = df[, col]
   sumry = summary(tmp)
   nms = names(sumry)
   nms = c(nms, 'std')
   out = c(sumry, sd(tmp))
   names(out) = nms
   out
   }
   describe(dat, 'horsepower')

   results including median, Q1(25%), Q2(50%), Q3(75%) and std. 

4. Visualize the statistics: 

   options(repos = c(CRAN = "http://cran.rstudio.com"))
   install.packages('gridExtra')
   plotstats = function(df, col, bins = 30){
   require(ggplot2)
   require(gridExtra)
   dat = as.factor('')
   ## Compute bin width
   bin.width = (max(df[, col]) - min(df[, col]))/ bins
   ## Plot a histogram
   p1 = ggplot(df, aes_string(col)) +
   geom_histogram(binwidth = bin.width)
   ## A simple boxplot
   p2 = ggplot(df, aes_string(dat, col)) +
   geom_boxplot() + coord_flip() + ylab('')
   ## Now stack the plots
   grid.arrange(p2, p1, nrow = 2)
   }
   
   plotstats(dat, 'price')

   results: