Practice: Data Exploration
Topic 1: Reading Histograms & Box Plots
SUPPORT Study: Background
 SUPPORT = Study to Understand Prognoses and Preferences for Outcomes and Risks of
Treatment
 4301 hospitalized patients from 5 different academic hospitals enrolled for phase 1 of the
study; 4028 patients were enrolled for phase 2.
 Overall goal was to develop a model that would estimate the overall survival for patients
based on their diagnosis and severity of illness.
** Main article for this study is cited on last slide and is posted in Collab (Resources > Data Sets >
SUPPORT Study) for those interested in reading about this study
Histograms & Boxplots - How to Describe
For Histograms, we look at:
1. Distribution/Shape
•
Symmetric or Skewed
•
How many peaks? (Unimodal, bimodal, or
multimodal)
•
Overall distribution (normal, uniform, etc)
2. Center
•
Median, mean
3. Variation/Spread
•
Observe min/max, large or small amount of
variation (standard deviation)
4. Potential Outliers?
For Boxplots, we look at:
1. 5 number summary
•
The five main values used to build the box
plot
2. Shape
•
Symmetric or skewed (look at the lengths
of the whiskers)
3. Outliers
•
Check with the outlier rule; R
automatically checks for outliers and will
use points to show outliers (if their are any)
White Blood Cell Counts - Exploration
Descriptive Statistics for
White Blood Cell Count
Min
0.050
Q1
6.899
Median
10.450
Q3
15.500
Max
100.000
Mean
12.400
Std Dev
8.9525
n = 1000
24 missing observations
Respiratory Rate - Exploration
Descriptive Statistics for
Respiratory Rate
Min
0.00
Q1
18.00
Median
24.00
Q3
29.00
Max
64.00
Mean
23.49
Std Dev
9.2557
n = 1000
No missing observations
 Data obtained from: http://biostat.mc.vanderbilt.edu/wiki/Main/DataSets
 Variable Information:
http://biostat.mc.vanderbilt.edu/wiki/pub/Main/DataSets/Csupport.html
 Article: Knaus WA, Harrell FE, Lynn J et al. (1995): The SUPPORT prognostic model: Objective
estimates of survival for seriously ill hospitalized adults. Annals of Internal Medicine 122:191203.