About

The focus of this lab is on data outliers, data preparation, and data modeling. This lab requires the use of Microsoft Excel, R, and ERDplus.

Setup

Remember to always set your working directory to the source file location. Go to ‘Session’, scroll down to ‘Set Working Directory’, and click ‘To Source File Location’. Read carefully the below and follow the instructions to complete the tasks and answer any questions. Submit your work to RPubs as detailed in previous notes.

Important Note

For your assignment you may be using different data sets than what is included in this worksheet demo. Make sure to read carefully the instructions on Sakai.

Task 1: Data Outliers

First, we must calculate the mean, standard deviation, maximum, and minimum for the Age column using R.

In R, we must read in the file again, extract the column and find the values that are asked for.

#Read File
mydata = read.csv(file="data/scoring.csv") 

#Name the extracted variable
age = mydata$Age 
#Calculate the average age below. Refer to Worksheet 2 for the correct command.
agemean=mean(age)
mean(age)
## [1] 37.08412
#Calculate standard deviation of age below. Refer to Worksheet 2 for the correct command. 
agesd=sd(age)
sd(age)
## [1] 10.98637
#Calculate the maximum of age below. The command to find the maximum is max(variable) where variable is the extracted variable.  
agemax=max(age)
max(age)
## [1] 68
#Calculate the minimum of age below. The command to find the minimum is min(variable) where variable is the extracted variable.  
agemin=min(age)
min(age)
## [1] 18

Next, use the formula from class to detect any outliers. An outlier is value that “lies outside” most of the other values in a set of data. A common way to estimate the upper and lower threshold is to take the mean (+ or -) 3 * standard deviation. Try using this formula to find the upper and lower limit for age.

#Use the formula above to calculate the upper and lower threshold
upper=agemean+3*agesd
lower=agemean-3*agesd
upper
## [1] 70.04322
lower
## [1] 4.125023

Another similar method to find the upper and lower thresholds discussed in introductory statistics courses involves finding the interquartile range. Follow along below to see how we first calculate the interquartile range..

quantile(age) 
##   0%  25%  50%  75% 100% 
##   18   28   36   45   68
lowerq = quantile(age)[2]
upperq = quantile(age)[4]
iqr = upperq - lowerq
lowerq
## 25% 
##  28
upperq
## 75% 
##  45
iqr
## 75% 
##  17

The formula below calculates the threshold. The threshold is the boundaries that determine if a value is an outlier. If the value falls above the upper threshold or below the lower threshold, it is an outlier.

Below is the upper threshold:

upperthreshold = (iqr * 1.5) + upperq 
upperthreshold
##  75% 
## 70.5

Below is the lower threshold:

lowerthreshold = lowerq - (iqr * 1.5)
lowerthreshold
## 25% 
## 2.5

Are there any outliers? How many? It can also be useful to visualize the data using a box and whisker plot. The boxplot below supports the IQR we found of 15 and upper and lower threshold.

There are no outliers because the lower quantile is above the lower threshold (2.5) and the higher quartile is below the upper threshold (70.5), meaning that our data falls within the range of the two calculated thresholds.

age[age > upperthreshold]
## integer(0)
age[age < lowerthreshold]
## integer(0)
boxplot(age, horizontal = TRUE)

Task 2: Data Preparation

Next, we must read the ‘creditriskorg.csv’ file into R. This is the original dataset and contains missing values.

newdata = read.csv(file="data/scoring_original.csv")
head(newdata)
##   Status Seniority  Home Time Age Marital Records       Job Expenses
## 1   good         9  rent   60  30 married  no_rec freelance     $73K
## 2   good        17  rent   60  58   widow  no_rec     fixed     $48K
## 3    bad        10 owner   36  46 married yes_rec freelance     $90K
## 4   good         0  rent   60  24  single  no_rec     fixed     $63K
## 5   good         0  rent   36  26  single  no_rec     fixed     $46K
## 6   good         1 owner   60  36 married  no_rec     fixed     $75K
##   Income Assets Debt     Amount      Price   Finrat   Savings
## 1  $129K      0    0   $800.00    $846.00  94.56265  4.200000
## 2  $131K      0    0 $1,000.00  $1,658.00  60.31363  4.980000
## 3  $200K   3000    0 $2,000.00  $2,985.00  67.00168  1.980000
## 4  $182K   2500    0   $900.00  $1,325.00  67.92453  7.933333
## 5  $107K      0    0   $310.00    $910.00  34.06593  7.083871
## 6  $214K   3500    0   $650.00  $1,645.00  39.51368 12.830769

We observe that the column names are shifted down below because of the empty line. So, we must make sure to use the command skip and set the header to true. (Not necessary for the scoring_original.cvs data because the column names are not shifted down below because of an empty line.)

To calculate the mean for Checking in R, follow Worksheet 2. Extract the Checking column first and then find the average using the function built in R. What happens when we try to use the function?

The functions results in an error, stating that the “argument is not numeric or logical.” This is because the data in the “Price” column contains dollar signs and commas, which are not numerical characters.

price = newdata$Price
pricemean=mean(price)
## Warning in mean.default(price): argument is not numeric or logical:
## returning NA

To resolve the error, we must understand where it is coming from and correct for. There are missing values in the csv file, which is quite common as most datasets are not perfect. Additionally, there are commas within the excel spreadsheet, and R does not recognize that ‘1,234’ is equivalent to ‘1234’. Lastly, there are ‘$’ symbols throughout the file which is not a numerical symbol either.

The sub function replaces these symbols with something else. So, in order to remove the comma in the number “1,234”, we must substitute it with just an empty space.

As shown on the worksheet, type and copy the exact commands to find the mean with the NA values removed.

#substitute comma with blank in all of checking.  Below are examples using a hypothetical variable name 'new'.
# Example new = sub(",","",new)
price = sub(",","",price)

#substitute dollar sign with blank in all of checking 
# Example new = sub("\\$","",new)
price = sub("\\$","",price)

#Convert values to numeric to remove any NA
# Example new = as.numeric(new)
price = as.numeric(price)
## Warning: NAs introduced by coercion
#Calculate mean of checking with NA removed 
# Example mean(new,na.rm=TRUE)
mean(price,na.rm=TRUE)
## [1] 1462.48

What are some other ways to clean this data? How about Excel? How does Excel treat the missing values and the “$” symbols?

Some other ways to clean this data are by analyzing the zeros and missing values, removing the decimals and deleting blanks (for example, some rows - especially towards the end - are empty and contain outliers -9999). Excel treats the missing values as if they were true 0s, which may be potentially wrong because the fact that a data is missing does not mean it can automatically be translated to a value of 0. The $ symbols denote currency. Nevertheless, they are disregarded and have no effect on the results of calculations. In other words, the simple helps to understand that we are dealing with dollars, but it is not taken into consideration when performing calculations.
————-

Task 3: Data Modeling

Now, we will look at Chicago taxi data. Go and explore the interactive dashboard and read the description of the data.

Chicago Taxi Dashboard: https://data.cityofchicago.org/Transportation/Taxi-Trips-Dashboard/spcw-brbq

Chicago Taxi Data Description: http://digital.cityofchicago.org/index.php/chicago-taxi-data-released/

– Open in RStudio or Excel the taxi trips sample csv file located in the data folder. Note the size of the file, the number of columns and of rows here. Identify the unique entities, and fields in the data.

newdata = read.csv(file="data/Taxi_Trips_Sample.csv")

Size of the file: 39.7 MB Number of columns: 24 Number of rows: 100,000 Unique entities: 100,000 Unique fields: 24

– Define a relational business logic integrity check for the column field ‘Trip Seconds’.

A relational business logic integrity in the column field ‘Trip Seconds’ can be found by creating a relation between time startstamp and time endstamp because they both determine the total time of the trips. Trip Seconds = Trip Time Endstamp - Trip Time Startstamp In other words, startstamp cannot be bigger (later in time) than endstamp, and seconds must equal to the difference between endstamp and startstamp.

– Using https://erdplus.com/#/standalone draw a star like schema using at least the following tables:

Star Schema

Star Schema