3/26/2011
1
Dispersion, Skewness and
Kurt osis
Kurt osis
Presen ted by: Mahen dra AN
Sources:
http:/ / busin ess.clayton .edu/ arjom an d/ busin ess/ stat%20 presen tation s/ busa310 1.htm l
Djarwan to, S ta tis tik S o s ia l Eko n o m iBagian pertam a edisi 3, BPFE, 20 0 1
Framework
Frequency distribution characteristics
Dispersion
•Dispersion is separatem easures of values am on g its cen tral
Absolut e Dispersion
Ra n ge
•Sim plest dispersion m easure
•Effected by extrem e values
•Effected by extrem e values
Qu a rtile D e via tio n
•Dispersion of in ter quartile ran ge
Absolut e Dispersion (cont . )
P e rce n tile D e via tio n
•Dispersion of in ter percen tile ran ge (P10an d
P90) P90)
Ave ra ge d e via tio n ( m e a n d e via tio n )
• Un grouped data
•Grouped data
Absolut e Dispersion (cont . )
S ta n d a rd d e via tio n
•Un grouped data
•Grouped data
3/26/2011
2
St andard Unit s
•To com pare two or m ore distribution s
•Stan dard un it show deviation of a variable value (X) on m ean ( ) in stan dard deviation un it (s) (X) on m ean ( ) in stan dard deviation un it (s)
•Com m on ly base on zero value (Z=0 )
•Exam ple: Z=1.2 is better than z=1.0
Relat ive Dispersion
•To kn ow sm allest variation in a distribution
•To com pare two or m ore frequen cy distribution
•All of stan dard dispersion m easurem en t can be
•All of stan dard dispersion m easurem en t can be used.
•Stated in coefficien t of variation
Skewness
•An im portan t m easure of the shape of a distribution is called skewn ess
•The form ula for com putin g skewn ess for a data
•The form ula for com putin g skewn ess for a data set is som ewhat com plex
Skewness (cont . )
Ka rl P e a rs o n m e th o d
•Base on m ean an d m edian values
B o w le y m e th o d
•Base on quartile values
Skewness (cont . )
10 – 9 0 p e rc e n tile 's m e th o d
•Base on percen tile
•The better m easurem en t for skewn ess base on the third m om en t.
Skewness (cont . )
•Grouped data skewn ess
•Third m om en t m ethod is used
3/26/2011
Distribution Shape:
Distribution Shape: SkewnessSkewness
SymmetricSymmetric((not skewed, SK = 0not skewed, SK = 0)) •
•If skewness is zero, thenIf skewness is zero, then
•
•Mean and median are equal.Mean and median are equal.
yy 3030
Distribution Shape:
Distribution Shape: SkewnessSkewness
Moderately Skewed LeftModerately Skewed Left •
•Is skewness is negative Is skewness is negative (left skewed SK >= (left skewed SK >= --3),3),thenthen
•
•Mean will usually be less than the median.Mean will usually be less than the median.
yy 3030
Distribution Shape:
Distribution Shape: SkewnessSkewness
Moderately Skewed RightModerately Skewed Right •
•If skewness is positive If skewness is positive (right skewed, SK >= +3),(right skewed, SK >= +3),thenthen
•
•Mean will usually be more than the median.Mean will usually be more than the median.
yy 3030
Distribution Shape:
Distribution Shape: SkewnessSkewness
Highly Skewed RightHighly Skewed Right •
•Skewness is positive (often above 1.0).Skewness is positive (often above 1.0).
•
•Mean will usually be more than the median.Mean will usually be more than the median.
yy 3030
•Measurem en t of peakedn ess
•Difficult to in terpret
•On ly in theoretic n ot in practice
•On ly in theoretic n ot in practice
Leptokurtic Mesokurtic
Plakurtic