Transcript Exploratory data analysis

Exploratory Data Analysis; coined by Tukey 1977
-Illuminate underlying pattern in noisy data
-Predecessor to formal analysis
-May lead to different analysis than originally planned
Data visualization
(The first thing you do with your data!!)
Important functions of exploratory data visualization
•Spot outliers
•Discriminate clusters
•Check distributional and other assumptions
•Examine relationships
•Compare mean differences
•Observe a time-based process
http://seamonkey.ed.asu.edu/~alex/teaching/WBI/EDA.html
Univariate data (one variable); frequency distributions
Distributions of height, biomass, etc…. often used to
describe populations
-How are the data distributed (including
summary/descriptive statistics)
-Are the data normal? (required to meet
assumptions of many statistical techniquesmore later)
-If not normal, can they be transformed?
Histograms
-Raw data hidden
-Division to categories arbitrary
-Excel, many programs
Identify outliers
Identify skew, non-normality
Stem-leaf plots
-show original data
-division to categories arbitrary
-easier to order data first
-a histogram on its side (sort of)
quiz scores
20
20
21
25
29
32
36
37
38
41
44
46
50
53
58
Stem
2
3
4
5
leaves
00159
2678
146
038
Box (box-whisker) plots
-calculate median, draw horizontal line
-draw a box with ends at the quartiles Q1 (25%) and Q3
(75%)
-extend the "whiskers" to the farthest points that are not
outliers
- outliers are outside 3/2 times the interquartile range (Q3Q1)
-Draw a dot for every outlier
Can be done for a
single distribution or
comparing several
http://mathworld.wolfram.com/Box-and-WhiskerPlot.html
Normal probability plots will be covered later
Bivariate (2 variable) data
-Relationship between the 2 variables
-Are there outliers?
-Examined by Scatterplots
negative
none
Non-linear
Graphing helps you see relationships. Formal analysis
guided by a priori knowledge that one variable causes
change in the other (more later)
Classified Data: often result from an ecological experiment
- Bar chart
-Shows means and variance
- “shows” treatment differences & magnitude
Epilithon NPP (mg O2/m2/hr)
15
10
5
0
-5
-10
high light
Mean  one S.E.
low light
List things that are wrong with this graph.
15
Epilithon NPP
10
5
0
-5
-10
Graphing Exercise
Obtain a dataset, preferably your own or a colleague’s,
but can be anything
Choose a graphing style that best illustrates the
“message” of your data
Use Excel or other program to make a graph
Print on an overhead to show in class