You are on page 1of 10

Dispersion parameters

Data 1 Data 2
 The concept of dispersion
20 0
40 30
50 50
60 70
80 100

 Range Mean 50 50
Dispersion parameters

Data 1 Data 2
Range 20 20
40 30
50 50
The Deviation from Mean 60 70
80 80
Mean 50 50
Solution: Variance and Deviation Range 60 60
Dispersion parameters

Data 1 Data 2
Variance and Deviation 20 20
we can consider positive deviation 40 30
50 50
weights and average them 60 70
80 80
Mean 50 50
But this measurement has a limitation Range 60 60
Dispersion parameters

Data 1 Data 2
Variance and Standard Deviation 20 20
has a different scale, so we can use absolute 40 30
50 50
deviation instead 60 70
80 80
Mean 50 50
Limitation: Compare {0,2,7} and {1,1,7} Range 60 60
Dispersion parameters

Data 1 Data 2
Standard Deviation 20 20
As MAD does not perform consistently when 40 30
50 50
small variations are observed, variance is 60 70
brought to the right scale. 80 80
Mean 50 50
Range 60 60
Limitation: higher computational cost
Dispersion parameters

 The Range
 Limitations and Solutions
 The Variance
 Limitations and Solutions
 The Standard Deviation
Dispersion parameters

Coef. of Variation (CV)


It is also computed as
Dispersion parameters

Chebyshev’s Theorem
For any data set, the percentage of observations that lie within k standard deviations of the
mean (i.e., within ) must be at least
The Empirical Rule
For data from a normal distribution, we expect the interval to contain a known percentage of
the data:
For
Outliers and standardized data

 Discard if the error is obvious


 In the 1980s, instruments monitoring the ozone layer over
Antarctica automatically disregarded readings two standard
deviations from the long-term mean as likely due to measurement
error (New Scientist, December 6, 2008, p. 32).
 Fortunately, the raw data were retrievable, and scientists were able
to spot an increase in the number of discarded readings as the
“ozone hole” grew.
Outliers and standardized data

Based on its standardized z-score, a data value is classified as:


Unusual if (beyond )
Outlier if (beyond )

You might also like