Summarizing Data

Question 1

Describe univariate data

Answer 1

• 1 variable
• For describing a distribution

Question 2

Describe bivariate and multivariate data

Answer 2

• 2 / 3 or more respectively
• For exploring relationships between variables

Question 3

Graphs that can be used for any type of categorical or quantitative data:

Answer 3

• Categorical:
  
  • Bar graph
• Quantitative:
  
  • Histogram
  • Dotplot
  • Boxplot

Question 4

Best graph for: 2 quantitative (and 1 categorical sometimes)

Answer 4

• Scatterplot

Question 5

Best graph for: 2 categorical

Answer 5

Two-way table

Question 6

Best graph for: 1 quantitative, 1 (or 2) categorical

Answer 6

• Stripchart
• Boxplot

Question 7

How do you determine skew?

Answer 7

• mean > median = right skew
  
  • Boxplot: mean is above the median
• mean < median = left skew
  
  • Boxplot: mean is below the median

Question 8

Define standard deviation

Answer 8

• Standard deviation: measure of spread about the mean

Question 9

Describe the interquartile ranges of a boxplot

Answer 9

• Each line is 25%
  
  • However, we cannot define any other percentage (ex., 85th percentile)
• Q1: median of numbers to the left
• Q2: median
• Q3: median of numbers to the right

Question 10

What is the 5 number summary?

Answer 10

Minimum Q1 M Q3 Maximum

Question 11

What is IQR and what is it used for?

Answer 11

• IQR = Q3 - Q1
• 1.5xIQR = outliers

Question 12

Define missing completely at random (MCAR)

Answer 12

• MCAR: missing value is truly random from the population (accidentally skipped the question, accidentally dropped the test tube)

Question 13

Define missing not at random (MNAR)

Answer 13

• MNAR: likelihood of missingness is associated with particular values (low test shore = higher chance of it missing)

Question 14

Define missing at random (MAR)

Answer 14

• MAR: a second variable influences the likelihood of missingness, but not the variable itself (females not reporting weight)

Question 15

Describe the 3 ways to handle missing values

Answer 15

1. Deletion
   
   • ​Complete case: get rid of the individual, decrease sample size
   • Pairwise: e.g. 8 individuals, only 7 data points for income
2. Imputation: replace with the mean/most frequent/predicted value (controversial)
3. Treat as a new category (e.g. a choice of colour is now ‘N/A’)

Question 16

What is metadata?

Answer 16

• Metadata: descriptive paragraph, table of information, or other files that help understand the data
  
  • Description of data collection
  • Where to access data
  • Known problems
  • Quality check characteristics (number of rows/columns, sum for numerical columns)