Non-Parametric Version of correlation coefficient

 

Spearman’s Rank Correlation

The concept of rank is used when the actual measurement or accurate assessment is not possible or the available information is qualitative in nature. In such a situation, the correlation failed, and an improvised statistical measure was developed to measure the strength and direction of association between two qualitative variables. 

For example, two persons are assigned to rank the 6 fruit juices (1 denotes first, the best taste).


The spearman’s rank correlation measures the strength and direction of association between two ranked variables or qualitative variables, which cannot be quantified in numerical form. This method was developed by Charles Edward Spearman in 1904.

The spearman rank correlation is used when the following assumptions are not met.

i. The data is not an interval or ratio.

ii. The data are not linearly related.

iii. The data is not bivariately distributed.

The spearman rank correlation between two sets of ranks is given by:


Practice Question

The two professors assign marks to five students on the basis of past experience in the two subjects, which are given below:

Find the coefficient of rank correlation.

Solution: The marks are not actual, so it is better to find spearman rank correlation instead of spearman correlation.

Step 1: Assign ranks


Step 2: Replace each observation by their respective rank.

Practice Question

Compute the rank correlation coefficient for the following data of the marks obtained by 8 students in statistics and mathematics.

Solution: Arrange the of both subjects and assigned ranks. 

Replace each observation by their respective rank.



In statistics (X), 23 is repeated 2 times, so m = 2



Kendall’s Tau Coefficient

Kendall’s Tau correlation coefficient is a statistical measure that assesses the strength of association between ranked variables. The Kendall tau rank correlation is a non-parametric alternative to the coefficient of Pearson correlation.

The Kendall coefficient can be calculated as:

Procedure to Perform Kendall’s test

Step 1: Assign ranks to both variables (X & Y).

Step 2: Arrange the ranks of X (column - X) in ascending order (from least to greatest) and rank of Y as it is.

Step 3: Count the number of concordant pairs (i.e., C) and discordant pairs (i.e., D). The sum of concordant pairs (C) & discordant pairs (D).

Step 4:  Use the following formula to calculate  

The large Tau value suggests a strong association.

The test of significance is given by:


Practice Question

The two experts ranked the quality of watermelon’s as follows:


Test the association by Kendall’s coefficient.

Solution:

Step 1: Assign ranks


Step 2: The rankings for expert 1 should be in ascending order.


Step 3: Find the concordant & disconcordant.


Kendall’s W Coefficient of Concordance

The Kendall's  test the degree of association between two sets of ranks. Now if the data is ranked by more than two judges, then the coefficient of concordance is used. The Kendall’s W statistic is given by:


The value of the W statistic lies between 0 & 1.

If W = 0, that means the judge ranked the list differently (or randomly).

If the W = 1, then everyone ranked the list in exactly the same order.

Where:

Practice Question

Given data


Find the coefficient of concordance and test the agreements between 3 sets of ranks.

Solution:







No comments:

Post a Comment

Moving Average Models (MA Models) Lecture 17

  Moving Average Models  (MA Models)  Lecture 17 The autoregressive model in which the current value 'yt' of the dependent variable ...