SlideShare une entreprise Scribd logo
1  sur  3
Télécharger pour lire hors ligne
03-09-2012




                                                                                   About Non-parametric (NP) Methods

                                                                             • Valid without restrictive assumptions about
                                                                               the population
           Nonparametric methods                                             • Rank based (detailed data not available or
                                                                               used)
                                                                             • More robust
                                                                                 – against outlier
                                                                                 – arbitrary population distribution
                              Session XIX                                    • Less efficient




                     A List of NP Methods
                                                                                        Tests for central tendency

                                                                           One-sample        Two-sample           K-sample
     •   Kolmogorov-Smirnov test
     •   Sign test
     •   Mann-Whitney U test/Rank-sum test                                 One-sample       Two-sample
                                                                           T-/Z- test for   T-/Z- test for        ANOVA
     •   Run test: A test for independence                                    mean          Equality of mean
     •   Kruskal-Wallis test


                                                                             Sign-test                             Kruskal-      Non-Parametric
                                                                                             Rank sum test
          What are their respective parametric counter-parts?               For median                            Wallis test
                                                                                             (Mann-Whitney)




                               Problem 1                                                     Problem 1: solution
A light-aircraft engine repair shop switched the payment method it used
from hourly wage to hourly wage plus a bonus computed on the time
                                                                              Change= before - after. +ve value indicates improvement
required to disassemble, repair, and reassemble an engine. The
following are data collected for 25 engines before the change and (and       H0: median(change) = 0 vs H1: median(change) > 0
the same) 25 engines after the change. At a 0.10 significance level, did
the new plan increase productivity?                                          H0: π=0.5 vs. H1: π > 0.5, where π =P(change > 0)
                  Hours Required         Hours Required
                Before       After     Before       After
                -------------------    -------------------
                   29         32          25         34
                   32         19          42         27
                   32         22          20         26
                   19         21          25         25
                   31         20          33         31
                   22         24          34         19
                   28         25          20         22
                   31         31          21         32
                   32         18          22         31
                   44         22          45         30
                   41         24          43         29
                   23         26          31         20
                   34         41




                                                                                                                                                   1
03-09-2012




                                                                                 Problem 1: solution
                       Sign test                                           (Assuming ‘the same’ 25 engines)
                                                                 Change= before - after. +ve value indicates improvement
• For paired comparison of ‘center’(median)                     H0: median(change) = 0 vs H1: median(change) > 0
  – One-sample test for median
                                                                H0: π=0.5 vs. H1: π > 0.5, where π =P(change > 0)
• Similarly test for percentiles
  – H0: 1st quartile of score out of 60 is 20                                      p - 0.5
                                                             Test Statistic Z =
  – T.S. is Y= no. of scores < 20 has B(n, .25) under the                         0.5 × 0.5
                                                                                       n
    null hypothesis                                          C.R. at level α = 0.1 is Z > 1.28                                                                           Conclusion at 10% level: no
• Take p-value approach if n is small                                                      13                                                                            significant increase in
                                                                                              - 0.5
                                                             Observed value of the T.S. is 23        = .63                                                               productivity.
                                                                                           0.5 × 0.5
                                                                                              23




          Statistics and Sleeping                                                                                       Problem 2
                                                            A manufacturer of toys changed the type of plastic molding machines it
                                                            was using because a new one gave evidence of being more economical.
                                                            As the Christmas season began, however, productivity seemed
                                                            somewhat lower than last year. Because production records of the past
                                                            years were readily available, the production manager decided to
                                                            compare the monthly output for the 15 months when the old machines
                                                            were used and the 11 months of production so far this year. Records
                                                            show these output amounts with the old and new machines.
                                                                Monthly output in Units
                                                                -------------------------------------------------------------------------------------------------
                                                                          Old Machines                                                                                        New Machines
                                                                          -------------------------------------------------                               -----------------------------------
                                                                992                                             966                                   965                                   956
                                                                945                                             889                                   1054                                  900
                                                                938                                             972                                   912                                   938
                                                                1027                                            940                                   850
                                                                892                                             873                                   796
                                                                983                                             1016                                  911
                                                                1014                                            897                                   877
                                                                1258                                                                                  902




            Problem 2: solution                                                                 Mann-Whitney U-test
                                                            • For comparing averages of two populations
                                                            • Combine all data and rank them.
 H 0 : m old = m new    vs   H1 : m old > m new                 – Smallest observation is assigned rank 1; second smallest
                                                                  observation is ranked 2 …Highest observation is ranked n
                                                                – Give mid- rank in case of tie.
mold average (median) output with old machine               • If the average of the first population is smaller, then the
mnew average (median) output with new machine                 combined rank of observations from first populations
                                                              would be small
                                                            • Need sample sizes to be 10 or more for normal
                                                              approximation (after standardizing). Tables are available for
                                                              smaller sample.
                                                            • Also known as Wilcoxon rank sum test




                                                                                                                                                                                                          2
03-09-2012




                            Calculation                                                                   Problem 2
      R1: Total rank of observations drawn from population 1                  H 0 : m old = m new              vs       H1 : m old > m new
      R2: Total rank of observations drawn from population 2
                                                                                                         n2 (n2 + 1)                 11× 12
                                                                             Test Statistic U = n1n2 +               − R2 = 15 ×11 +        − R2 = 231 − R2
                                                                                                              2                        2
                               n1 ( n1 + 1)                                                           U - µU
Test Statistic U = n1n2 +                   − R1                             At α = 0.1, the C.R is            > 1.28
                                     2                                                                 σU
= No. of (X i , Y j ) pairs with X i < Y j ( R1 = rank sum of all X obs)     where, µU =
                                                                                           n1n2 165
                                                                                               =
                                                                                                           n n (n + n + 1)
                                                                                                    , σU = 1 2 1 2         = 19.27
                                                                                            2    2               12
                   n1n2             n1n2 (n1 + n2 + 1)
Under H 0 , µU =        ,    σU =
                    2                      12                                                                                (231 − 115.5) − 82.5
                                                                           R2=115.5, Observed value of the T.S. is                  19.27
                                                                                                                                                  = 1.71
      U-µ U
and         has N(0,1) distribution
       σU                                                                   So Reject H0 at 10% level and conclude that the change has reduced
                                                                            the average output level




                                                                                                                                                                   3

Contenu connexe

En vedette

Observatoire sur les grandes acquisitions foncières à Madagascar
Observatoire sur les grandes acquisitions foncières à MadagascarObservatoire sur les grandes acquisitions foncières à Madagascar
Observatoire sur les grandes acquisitions foncières à MadagascarIlc Landcoalition
 
Design logos
Design logosDesign logos
Design logosDoey19
 
Man org session 3-org and technology_5th july 2012
Man org session 3-org and technology_5th july 2012Man org session 3-org and technology_5th july 2012
Man org session 3-org and technology_5th july 2012vivek_shaw
 
Understanding students with disabilities (ch. 2) 2.13.13
Understanding students with disabilities (ch. 2) 2.13.13Understanding students with disabilities (ch. 2) 2.13.13
Understanding students with disabilities (ch. 2) 2.13.13adriewool
 
Teaching all students (ch.5) 3.6.13
Teaching all students (ch.5) 3.6.13Teaching all students (ch.5) 3.6.13
Teaching all students (ch.5) 3.6.13adriewool
 

En vedette (6)

Observatoire sur les grandes acquisitions foncières à Madagascar
Observatoire sur les grandes acquisitions foncières à MadagascarObservatoire sur les grandes acquisitions foncières à Madagascar
Observatoire sur les grandes acquisitions foncières à Madagascar
 
Design logos
Design logosDesign logos
Design logos
 
Session 17
Session 17Session 17
Session 17
 
Man org session 3-org and technology_5th july 2012
Man org session 3-org and technology_5th july 2012Man org session 3-org and technology_5th july 2012
Man org session 3-org and technology_5th july 2012
 
Understanding students with disabilities (ch. 2) 2.13.13
Understanding students with disabilities (ch. 2) 2.13.13Understanding students with disabilities (ch. 2) 2.13.13
Understanding students with disabilities (ch. 2) 2.13.13
 
Teaching all students (ch.5) 3.6.13
Teaching all students (ch.5) 3.6.13Teaching all students (ch.5) 3.6.13
Teaching all students (ch.5) 3.6.13
 

Plus de vivek_shaw

Lecture 11 market structure- perfect competition
Lecture 11  market structure- perfect competitionLecture 11  market structure- perfect competition
Lecture 11 market structure- perfect competitionvivek_shaw
 
Lecture 9 costs
Lecture 9  costsLecture 9  costs
Lecture 9 costsvivek_shaw
 
Lecture 8 production, optimal inputs
Lecture 8  production, optimal inputsLecture 8  production, optimal inputs
Lecture 8 production, optimal inputsvivek_shaw
 
Lecture 8 production, optimal inputs (1)
Lecture 8  production, optimal inputs (1)Lecture 8  production, optimal inputs (1)
Lecture 8 production, optimal inputs (1)vivek_shaw
 
Lecture 3 dds sand elasticity
Lecture 3  dds sand elasticityLecture 3  dds sand elasticity
Lecture 3 dds sand elasticityvivek_shaw
 
Consumertheory1
Consumertheory1Consumertheory1
Consumertheory1vivek_shaw
 
Consumer theory 2
Consumer theory 2Consumer theory 2
Consumer theory 2vivek_shaw
 
Class2 market, demand and supply
Class2  market, demand and supplyClass2  market, demand and supply
Class2 market, demand and supplyvivek_shaw
 
Policy implications
Policy implicationsPolicy implications
Policy implicationsvivek_shaw
 
Man org session 14_org decision making_16th august 2012
Man org session 14_org decision making_16th august 2012Man org session 14_org decision making_16th august 2012
Man org session 14_org decision making_16th august 2012vivek_shaw
 
Man org session 12_org learning_3rd august 2012
Man org session 12_org learning_3rd august 2012Man org session 12_org learning_3rd august 2012
Man org session 12_org learning_3rd august 2012vivek_shaw
 
Man org session 11 interorganizational relationships_2nd august 2012
Man org session 11 interorganizational relationships_2nd august 2012Man org session 11 interorganizational relationships_2nd august 2012
Man org session 11 interorganizational relationships_2nd august 2012vivek_shaw
 
Man org session 10_org control_27th july 2012
Man org session 10_org control_27th july 2012Man org session 10_org control_27th july 2012
Man org session 10_org control_27th july 2012vivek_shaw
 
Man org session 9_org culture_26th july 2012
Man org session 9_org culture_26th july 2012Man org session 9_org culture_26th july 2012
Man org session 9_org culture_26th july 2012vivek_shaw
 

Plus de vivek_shaw (20)

Lecture 11 market structure- perfect competition
Lecture 11  market structure- perfect competitionLecture 11  market structure- perfect competition
Lecture 11 market structure- perfect competition
 
Lecture 9 costs
Lecture 9  costsLecture 9  costs
Lecture 9 costs
 
Lecture 8 production, optimal inputs
Lecture 8  production, optimal inputsLecture 8  production, optimal inputs
Lecture 8 production, optimal inputs
 
Lecture 8 production, optimal inputs (1)
Lecture 8  production, optimal inputs (1)Lecture 8  production, optimal inputs (1)
Lecture 8 production, optimal inputs (1)
 
Lecture 3 dds sand elasticity
Lecture 3  dds sand elasticityLecture 3  dds sand elasticity
Lecture 3 dds sand elasticity
 
Game theory 3
Game theory 3Game theory 3
Game theory 3
 
Game theory 1
Game theory 1Game theory 1
Game theory 1
 
Game theory 2
Game theory 2Game theory 2
Game theory 2
 
Ford motors
Ford motorsFord motors
Ford motors
 
Consumertheory1
Consumertheory1Consumertheory1
Consumertheory1
 
Consumer theory 2
Consumer theory 2Consumer theory 2
Consumer theory 2
 
Class2 market, demand and supply
Class2  market, demand and supplyClass2  market, demand and supply
Class2 market, demand and supply
 
Auctions 1
Auctions 1Auctions 1
Auctions 1
 
Costs2
Costs2Costs2
Costs2
 
Policy implications
Policy implicationsPolicy implications
Policy implications
 
Man org session 14_org decision making_16th august 2012
Man org session 14_org decision making_16th august 2012Man org session 14_org decision making_16th august 2012
Man org session 14_org decision making_16th august 2012
 
Man org session 12_org learning_3rd august 2012
Man org session 12_org learning_3rd august 2012Man org session 12_org learning_3rd august 2012
Man org session 12_org learning_3rd august 2012
 
Man org session 11 interorganizational relationships_2nd august 2012
Man org session 11 interorganizational relationships_2nd august 2012Man org session 11 interorganizational relationships_2nd august 2012
Man org session 11 interorganizational relationships_2nd august 2012
 
Man org session 10_org control_27th july 2012
Man org session 10_org control_27th july 2012Man org session 10_org control_27th july 2012
Man org session 10_org control_27th july 2012
 
Man org session 9_org culture_26th july 2012
Man org session 9_org culture_26th july 2012Man org session 9_org culture_26th july 2012
Man org session 9_org culture_26th july 2012
 

Session 19

  • 1. 03-09-2012 About Non-parametric (NP) Methods • Valid without restrictive assumptions about the population Nonparametric methods • Rank based (detailed data not available or used) • More robust – against outlier – arbitrary population distribution Session XIX • Less efficient A List of NP Methods Tests for central tendency One-sample Two-sample K-sample • Kolmogorov-Smirnov test • Sign test • Mann-Whitney U test/Rank-sum test One-sample Two-sample T-/Z- test for T-/Z- test for ANOVA • Run test: A test for independence mean Equality of mean • Kruskal-Wallis test Sign-test Kruskal- Non-Parametric Rank sum test What are their respective parametric counter-parts? For median Wallis test (Mann-Whitney) Problem 1 Problem 1: solution A light-aircraft engine repair shop switched the payment method it used from hourly wage to hourly wage plus a bonus computed on the time Change= before - after. +ve value indicates improvement required to disassemble, repair, and reassemble an engine. The following are data collected for 25 engines before the change and (and H0: median(change) = 0 vs H1: median(change) > 0 the same) 25 engines after the change. At a 0.10 significance level, did the new plan increase productivity? H0: π=0.5 vs. H1: π > 0.5, where π =P(change > 0) Hours Required Hours Required Before After Before After ------------------- ------------------- 29 32 25 34 32 19 42 27 32 22 20 26 19 21 25 25 31 20 33 31 22 24 34 19 28 25 20 22 31 31 21 32 32 18 22 31 44 22 45 30 41 24 43 29 23 26 31 20 34 41 1
  • 2. 03-09-2012 Problem 1: solution Sign test (Assuming ‘the same’ 25 engines) Change= before - after. +ve value indicates improvement • For paired comparison of ‘center’(median) H0: median(change) = 0 vs H1: median(change) > 0 – One-sample test for median H0: π=0.5 vs. H1: π > 0.5, where π =P(change > 0) • Similarly test for percentiles – H0: 1st quartile of score out of 60 is 20 p - 0.5 Test Statistic Z = – T.S. is Y= no. of scores < 20 has B(n, .25) under the 0.5 × 0.5 n null hypothesis C.R. at level α = 0.1 is Z > 1.28 Conclusion at 10% level: no • Take p-value approach if n is small 13 significant increase in - 0.5 Observed value of the T.S. is 23 = .63 productivity. 0.5 × 0.5 23 Statistics and Sleeping Problem 2 A manufacturer of toys changed the type of plastic molding machines it was using because a new one gave evidence of being more economical. As the Christmas season began, however, productivity seemed somewhat lower than last year. Because production records of the past years were readily available, the production manager decided to compare the monthly output for the 15 months when the old machines were used and the 11 months of production so far this year. Records show these output amounts with the old and new machines. Monthly output in Units ------------------------------------------------------------------------------------------------- Old Machines New Machines ------------------------------------------------- ----------------------------------- 992 966 965 956 945 889 1054 900 938 972 912 938 1027 940 850 892 873 796 983 1016 911 1014 897 877 1258 902 Problem 2: solution Mann-Whitney U-test • For comparing averages of two populations • Combine all data and rank them. H 0 : m old = m new vs H1 : m old > m new – Smallest observation is assigned rank 1; second smallest observation is ranked 2 …Highest observation is ranked n – Give mid- rank in case of tie. mold average (median) output with old machine • If the average of the first population is smaller, then the mnew average (median) output with new machine combined rank of observations from first populations would be small • Need sample sizes to be 10 or more for normal approximation (after standardizing). Tables are available for smaller sample. • Also known as Wilcoxon rank sum test 2
  • 3. 03-09-2012 Calculation Problem 2 R1: Total rank of observations drawn from population 1 H 0 : m old = m new vs H1 : m old > m new R2: Total rank of observations drawn from population 2 n2 (n2 + 1) 11× 12 Test Statistic U = n1n2 + − R2 = 15 ×11 + − R2 = 231 − R2 2 2 n1 ( n1 + 1) U - µU Test Statistic U = n1n2 + − R1 At α = 0.1, the C.R is > 1.28 2 σU = No. of (X i , Y j ) pairs with X i < Y j ( R1 = rank sum of all X obs) where, µU = n1n2 165 = n n (n + n + 1) , σU = 1 2 1 2 = 19.27 2 2 12 n1n2 n1n2 (n1 + n2 + 1) Under H 0 , µU = , σU = 2 12 (231 − 115.5) − 82.5 R2=115.5, Observed value of the T.S. is 19.27 = 1.71 U-µ U and has N(0,1) distribution σU So Reject H0 at 10% level and conclude that the change has reduced the average output level 3