Non Parametric Methods

What are Non Parametric Methods?

A commonly used method in statistics where

small sample sizes are used to analyze nominal data. Nonparametric method is used when the researcher does not know anything about the parameters of the sample chosen from the population.
his method is sometimes referred to as parameter-free or

distribution-free method.

Types of Non-parametric Methods

he two commonly used non-parametric correlations are !
"pearman#s $ank %orrelation &endall#s au

Spearman's Rank Correlation

'n statistics( Spearman's rank correlation

or Spearman's rho( named after %harles "pearman and often denoted by the )reek letter *rho+ or as ( is a nonparametric measure of statistical dependence between two ,ariables.
't assesses how well the relationship between two

,ariables can be described using a monotonic function.

'f there are no repeated data ,alues( a perfect "pearman

correlation of -. or /. occurs when each of the ,ariables is a perfect monotone function of the other.

 wo sets of ,ariables are taken and ranks are assigned

according to their positions in the ascending order of the ,alues.

0ifferences between the ranks of each obser,ation on the two

,ariables are calculated( and 1 is gi,en by!

23ample !

 he standard error of the coefficient *+ was determined

by Pearson in .456 and )osset in .475. 't is

$esults
'f the ,alue is -.( there is a perfect negati,e correlation 'f the ,alue falls between -. and -5.8( there is a strong negati,e correlation 'f the ,alue falls between -5.8 and 5( there is a weak negati,e correlation 'f the ,alue is 5( there is no correlation 'f the ,alue falls between 5 and 5.8( there is a weak positi,e correlation 'f the ,alue falls between 5.8 and .( there is a strong positi,e correlation 'f the ,alue is .( there is a perfect positi,e correlation between the 7 sets of data. 'f the ,alue is 5( state that null hypothesis is accepted. 9therwise( say it is re:ected.

Kendalls Rank Correlation

he Kendalls rank correlation ( commonly referred to as Kendall's tau ( ( is a statistic method used to measure the association between two measured ;uantities. statistical dependence based on the tau coefficient.

A tau test is a non parametric hypothesis test for "pecifically( it is a measure of rank correlation( i.e.( the

similarity of the orderings of the data when ranked by each of the ;uantities. 't is named after Maurice &endall( who de,eloped it in .4<=.

he &endall > coefficient is defined as!

%oncordant pairs ? 0iscordant pairs %oncordant pairs - 0iscordant pairs

$ange!- -..5 to ..5

@ypothesis test
he &endall#s rank correlation is often used in a statistical

hypothesis test to establish whether two ,ariables may be regarded as statistically dependent. his test is nonparametric( as it does not rely on any assumptions on the distributions of X or Y or the distribution of *X(Y+.

!d"anta#es of Non-parametric Methods

Nonparametric methods re;uire no or ,ery limited

assumptions to be made about the format of the data( and they may therefore be preferable when the assumptions re;uired for parametric methods are not ,alid.
Nonparametric methods can be useful for dealing with

une3pected( outlying obser,ations that might be problematic with a parametric approach.

Nonparametric methods are intuiti,e and are simple to

carry out by hand( for small samples at least.

Nonparametric methods are often useful in the analysis of

ordered categorical data in which assignation of scores to indi,idual categories may be inappropriate. Aor e3ample( non-parametric methods can be used to analyse soft drink consumption directly using the categories ne,er( a few times per year( monthly( weekly( a few times per week( daily and a few times per day. 'n contrast( parametric methods re;uire scores *i.e. .?6+ to be assigned to each category.

$isad"anta#es of Non-parametric Methods

Nonparametric methods may lack power as compared with

more traditional approaches . his is a particular concern if the sample size is small or if the assumptions for the corresponding parametric method.
Nonparametric methods are geared toward hypothesis

testing rather than estimation of effects. 't is often possible to obtain nonparametric estimates and associated confidence inter,als( but this is not generally straightforward.

 ied ,alues can be problematic when these are common(

and ad:ustments to the test statistic may be necessary.

Appropriate computer software for nonparametric

methods can be limited( although the situation is impro,ing. 'n addition( how a software package deals with tied ,alues or how it obtains appropriate P ,alues may not always be ob,ious.

%onclusion
he use of non-parametric methods may be necessary when data ha,e a ranking

but no clear numerical interpretation( such as when assessing preferences.

Non-parametric methods make fewer assumptions( heir applicability is much wider than the corresponding parametric methods. 'n

particular( they may be applied in situations where less is known about the application in ;uestion. more robust.

Also( due to the reliance on fewer assumptions( non-parametric methods are he use of non-parametric methods is simple and easy. But the wider applicability and increased robustness of non-parametric tests comes

at a cost! in cases where a parametric test would be appropriate( non-parametric tests ha,e less power( Aor e3ample. larger sample size.

Find the correlation between the price of brinjals and onions.

Year 1988 1989 1990 1991 1992 199 1994 1995 1996 1997 1998 1999 2000 Brinjal Prices (Rs/1kg) 66.60 69.50 74.60 72.70 71. 0 72.60 66.70 61.80 58.70 6 .10 59.60 6 .40 68.60 Onion Prices (Rs/1kg) 69.10 66.10 55.50 52.20 59.50 64.40 65.60 78.20 82.80 90. 0 72. 0 74.50 79.40

Year 1988 1989 1990 1991 1992 199 1994 1995 1996 1997 1998 1999 2000

Brinjal Prices (Rs/1kg) 66.60 69.50 74.60 72.70 71. 0 72.60 66.70 61.80 58.70 6 .10 59.60 6 .40 68.60

Onion Prices (Rs/1kg) 69.10 66.10 55.50 52.20 59.50 64.40 65.60 78.20 82.80 90. 0 72. 0 74.50 79.40

Rank! Brinjal 6 9 1 12 10 11 7 1 4 2 5 8

Rank! Onion 7 6 2 1 4 5 10 12 1 8 9 11

" #1 11 11 7 7 2 #7 #11 #9 #6 #4 #

"2 1 9 121 121 49 49 4 49 121 81 6 16 9 666