inst/examples/workflow-part2.R

In bfatemi/easydata: Utility functions for easy data wrangling

# # Now we want make our work reproducible and 
# # apply it to the rest of the similarly structured tables
# 
# # let's functionalize the code we wrote and make minor tweaks for reusability
# 
# # give function a helpful name. Refer to a style guide or make your own, but be consistent
# 
# #cleanmelt for "clean" and "melt" which is what our function does
# CleanMelt <- function(dt){
#     # make all integer and logical -> numeric
#     ClassMorph(dt,"logical","numeric")
#     ClassMorph(dt,"integer","numeric")
#     
#     # Demonstrate cRemove
#     
#     # remove columns where ALL values are NA (last column)
#     cRemoveNA(dt) 
#     
#     # make table long-form
#     dtlong <- melt(dt, id.vars = "COUNTRY", 
#                    variable.name = "Quarter")
#     
#     # because the new column "Quarter" was a mix of characters and numbers seperated by a space,
#     # R converted them to factors. Convert all factors to characters
#     ClassMorph(dtlong, "factor", "character")
#     
#     # Notice that column quarter also has year. lets split that up
#     dtlong[, (c("Quarter", "Year")):=(as.data.table(do.call(rbind,strsplit(Quarter, " "))))]
#     
#     return(dtlong)
# }
# 
# # place all of our functions in a list and use "lapply" to apply it
# # good practice to name the list if we are holding similar structures with otherwise no
# # distinguishing attributes
# #
# llraw <- list(AveSpeed_Mbps = dt_ave, 
#                PeakSpeed_Mbps = dt_peak, 
#                Access_Broadband = dt_bb, 
#                Access_HighBroadband = dt_hbb, 
#                Access_LowBroadband = dt_nbb)
# ll <- lapply(llraw, CleanMelt)
# 
# # Notice how we have a column called "value" for each? Lets be more specific like our script,
# # however we couldn't generalize it in our function because each table's "value" column should
# # have a different name. Let's use the names we gave our data in the list as it describes the
# # entries.
# 
# # write a simple loop to change a dt's column name from "value" to a given input
#         # Note how I access elements of "llres" by name
# for(i in names(ll))
#     setnames(ll[[i]], "value", i)
# 
# #-----------------------------------------
# # MORPH DATA TO DESIRED FORM
# #-----------------------------------------
# 
# # Keep as list seperated by measure variable type
# ll
# 
# # Merge them together into a wider table
# wideDT <- Reduce(merge, ll)
# 
# # Melt the wider table and add a column identifying the type of measure
# longDT <- melt(DT, id.vars = c("COUNTRY", "Quarter", "Year"))
# 
# #-----------------------------------------
# # EXPLORE THE DATA (IN VARIOUS FORMS)
# #-----------------------------------------
# 
# # Demonstrate rFilter
# 
# # Since the data is in "long" form (much better for analysis), we need to be
# # aware of a key nuance: in our new "variable" column, we have two types of
# # measures. We have "Access" (a percentage), and "Speed" (Mbps). 
# #
# # filter rows for "Access" measures (does a fast text search accross columns)
# rFilter(longDT, val = "Access")
# 
# # Lets say we quickly want to quickly create a new column that identifies
# # whether the row corresponds to a percentage measure or a speed measure:
# a <- rFilter(longDT, val = "Access", set = list(IsAccess=1))
# b <- rFilter(longDT, val = "Speed", set = list(IsAccess=0))
# 
# longDT <- rbindlist(list(a,b))
# 
# # Demonstrate xSplit
# 
# # Let's take a quick look all data points such that access was 
# # above 50 in 09 
# splice(DT=longDT, 
#        Year = "09",
#        Quarter = "Q1" & "Q4",
#        Access = 1, 
#        value > 50)
# 
# # Two new areas in Q4 and also Delaware access grew by 5% in 2009
# 
# # Let's try the same thing on the wide view. This time, we can apply
# # a statistics function to a different column after the splice:
# 
# # For the same cut of data answer this question:
# 
# # What is the average mbps speed in places where broadband access is 75% at 
# # the beginning and end of 2011
# splice(DT = wideDT,
#        Quarter = "Q1" & "Q4",
#        Year = "11",
#        Access_Broadband > 75,
#        apply = mean(AveSpeed_Mbps))
# 
# # Could there be a relationship between access and speed? Makes sense:
# # the wider availability could bring more competition among internet providers
# # to offer higher speeds. I'm assuming a lot so entertain me by not pointing
# # out the shoddy conclusions
# 
# # Let's investigate just a bit further and see if the same thing is happeing 
# # as a general trend aggregating all years:
# splice(DT = wideDT,
#        Quarter = "Q1" & "Q4",
#        Access_Broadband > 75,
#        apply = mean(AveSpeed_Mbps))
# 
# # How about between years (aggregating the quarters)
# splice(DT = wideDT,
#        Year = "10" & "11" & "12" & "13",
#        Access_Broadband > 75,
#        apply = mean(AveSpeed_Mbps))