R语言初步(一）

Assignment 1

Description: location choice of Belgium ATMs
Data: Belgium ATM distribution in 1994

预处理

# clear everything
rm(list = ls())
cat("\014")   # clear console

# load data
# url <- 'https://dl.dropboxusercontent.com/s/q6qzbfa1tdcqv6v/belgium_atm.csv'
url <- "D://R//code//belgium_atm.csv"
# 换成本地执行路劲
df <- read.csv(url, stringsAsFactors = F)

# we can check the structure of the data by running

head(df)

> head(df)
  population numATMs   ATMwithdr withdrvalue unemprate numbranches
1       3722       1   .25542593 79.13402557 0.0728676       0.500
2       7006       2 1.837865114 102.6663437 0.0226948       0.500
3       4234       0     missing     missing 0.0273973       0.125
4       6229       0     missing     missing 0.0244020       0.750
5      10303       1 .6062539816 98.93833923 0.0284383       0.375
6       7424       0     missing     missing 0.0373114       0.875

Q1

# ==== question 1 ====

# Q1. First, recall that df is a data frame which is like a spreadsheet in Excel. 
#   Let's convert every column into a separate variable using '$'; for example:

population <- df$population
numATMs <- df$numATMs
ATMwithdr = df$ATMwithdr
withdrvalue = df$withdrvalue
unemprate = df$unemprate
numbranches = df$numbranches

# do the same for the other columns

# ------ you should not work with 'df' anywhere beyond this line -----
#   i.e. please only work with the vectors (or create new vectors)
#   for the rest of this assignment

Q2

# ==== question 2 ====

# Q2a. Do the necessary conversion for all variables so that you can apply numeric operations on them
#   replace the original value of the vector in case of conversion

# dataframe
str(df)

df$ATMwithdr = as.numeric(df$ATMwithdr)
df$withdrvalue = as.numeric(df$withdrvalue)

# Q2b. population is in a very different scale. Rescale it into thousands, i.e., divide population by 1000
#   and replace the variable

df$population = df$population / 1000

Q3

# ==== question 3 ====

# You want to take average for all variables but you realized that some variables have missing value
#   before taking averages, you need to make sure that all observations are taken from the same sets of 
#   observations (i.e. rows) where no variable is missing 

# Q3a. let's define a logical vector for non-missing rows, i.e. rows without any missing values, name it 'nm'
#   note: nm will be a vector which is the same length as the number of rows in the original data df

# check which column has missing values
nm = !(is.na(df$ATMwithdr) | is.na(df$withdrvalue))

# Q3b. count the number of non-missing rows in the data df, name it 'count_nm'
#   count_nm should be one number

count_nm = length(nm[nm == T])

# T = 1, F = 0
count_nm = sum(nm)

# dim: row, column
count_nm = dim(df[nm,])[1]

Q4

# ==== question 4 ====

# Q4. Calculate the averages of number of ATM, number of branches, population, 
#   unemployment rate, number of withdraw per resident and amount per withdrawl.
#   In particular, notice that certain variables have missing values and you might want to  
#   only calculate means for the rows without missing values of any variable
#   (that is, the rows that you use to calculate the average of all variables should be the same)
#   Finally, collect results in a vector called 'mean_nm', name elements in the vector by the original variable name

mean(df$population[nm])
mean(df[nm, 'population'])
mean_nm = c(population = mean(df[nm, 'population']), numATMs = mean(df[nm, 'numATMs']), 
            ATMwithdr = mean(df[nm, 'ATMwithdr']), withdrvalue = mean(df[nm, 'withdrvalue']), 
            unemprate = mean(df[nm, 'unemprate']), numbranches = mean(df[nm, 'numbranches']))
mean_nm

> count_nm = dim(df[nm,])[1]
> mean(df$population[nm])
[1] 0.01344537
> mean(df[nm, 'population'])
[1] 0.01344537
> mean_nm
  population      numATMs    ATMwithdr  withdrvalue    unemprate  numbranches 
  0.01344537   1.56774194   0.86759136 100.52755132   0.03167726   1.24960350 

Q5

# ==== question 5 ====

# Q5. You realize that the reason for missing values in the original data is that there are no ATMs.
#   So in that regard you could have defined the missing values to zero
#   Re-define the missings to zero and assign it to the original variable,
#   find the total number of observations in the dataset (call it 'count_all'), 
#   and re-calculate means for the same set of variables and collect results in 'mean_all'

df[is.na(df)] = 0
count_all = dim(df)[1]
mean_all = c(population = mean(df$population), numATMs = mean(df$numATMs), 
            ATMwithdr = mean(df$ATMwithdr), withdrvalue = mean(df$withdrvalue), 
            unemprate = mean(df$unemprate), numbranches = mean(df$numbranches))
mean_all
> mean_all
  population      numATMs    ATMwithdr  withdrvalue    unemprate  numbranches 
 0.008738464  0.737481032  0.408123401 47.289136432  0.030950722  0.856129873 

Q6

# ==== question 6 ====

# You decide to investigate what's the average number of withdrawal and amount per withdrawal
#   by areas with different number of ATMs

# Q6a. Let's summarize average ATMwithdr and average withdrvalue by the number of atms (for range 1-4). 
#   collect results in two separate vectors and name them 'mean_a' and 'mean_w'

mean_a = c(mean(df[df$numATMs == 1, 'ATMwithdr']), mean(df[df$numATMs == 2, 'ATMwithdr']),
           mean(df[df$numATMs == 3, 'ATMwithdr']), mean(df[df$numATMs == 4, 'ATMwithdr']))
mean_a

mean_w = c(mean(df[df$numATMs == 1, 'withdrvalue']), mean(df[df$numATMs == 2, 'withdrvalue']),
           mean(df[df$numATMs == 3, 'withdrvalue']), mean(df[df$numATMs == 4, 'withdrvalue']))
mean_w

> mean_w
[1] 101.03531 100.37617  99.25264  96.25565

# Q6b. Separately, plot these by the number of ATMs; label the x axis "number of ATMs" and y axis 
#   "average withdrawl per resident" and "average amount per withdrawl", respectively
#   use line plot by setting type = 'l' as one of the plot function arguments
plot(c(1, 2, 3, 4), mean_a, type = 'l', xlab = 'number of ATMs', 
     ylab = 'average withdrawl per resident')
plot(c(1, 2, 3, 4), mean_w, type = 'l', xlab = 'number of ATMs', 
     ylab = 'average withdrawl per resident')