Are kendall tau b, Spearman's rho, Gamma also tell us the significance or should it only be used to refer to direction and strength of the relationship. In SPSS it shows the P value (approx sig in Chi Square) and Sig 2 value in Bivariate correlation in SPSS. I am having trouble running Chi Square for finding test of significance due to expected counts. Can I only refer to Kendall Tau b, Spearman rho or Gamma for the direction, strength and significance of the relationship/association and ignore Chi Square. Also can you let me now the alternative to Chi square to test the significance of the association/relationship between the two variables please
Thanks
Used to study the relationship and the strength of the relationship between variables
Hello Amit,
To answer the last part of your query first, it depends on what the two variables are, how they are quantified, their distribution, and the definition of association you'd like to apply as to what best specific option for testing association might be.
If you're having difficulty eliciting chi-square tests (for what I presume must be categorical data), then the usual culprit is empty or near-empty cells (which often associates with having one or more instances of expected cell frequencies less than 5).
Kendall's tau, Spearman's rho, and Goodman/Kruskal's gamma all can be evaluated for statistical significance (difference from zero) with an exact calculation of the p-value. [Note that these all may be applied to ordinal data, but each defines association differently.] Many statistical packages, however, use "large-sample" approximations to determine probability of an observed relationship (under the null hypothesis) because it's easier to program one solution for p-value than to build in look-up tables or compute exact probabilities for large sample sizes.
Good luck with your work!
Thank you very much for your response David Morse .
I am considering two likert type statements which are getting measured on a 5 point likert scale (SA-1,Somewhat A-2,Neutral-3,Somewhat Disagree-4,Strongly Disagree-5). For example Statement 1 is "I prefer to have access to social networking sites at the workplace" and Statement 2-"I consider using of social networking sites at the workplace as a distraction". These both are ordinal variables (I am considering both these statements as independent of each other/i.e no relationship though one can also lead to another?). Like this there are many Likert items in my questionnaire for which I have run a correlation matrix analysis (Bivariate) to check if there is any association between these statements. I did this to get some cues from my data. I have run Kendell Tau b for this. The cross tabulation was for 2 ordinal variables and was having 5 by 5 table (I have removed " do not with to answer/not applicable" response categories before running Kendell tau b) which has now been grouped and converted to 3 point scale (Agree-1, Neutral-2, Disagree-3) so that to improve the expected counts for Chi Square but still it is giving issue as the sample size is 127 and for some variables based on the removal of "do not wish to answer" and "N/a as I am not employed" categories goes down to 107 and 111.
* [Note that these all may be applied to ordinal data, but each defines association differently.] - Could you please share more knowledge about these measures of association in terms of the association that they define (differently). From this statement, I am considering that they should be used based on what I want to find between the two variables/statements.
From the literature review, it came to light that good percentage of Gen Zs who like to have access to social networking sites at the workplace also consider it to be a distraction. That is why I wanted to check the association between these two statements. The result that I have got from Kendell Tau b (used for equal no. of rows and columns) for statement 1- " like to have access to social networking sites at the workplace" and statement 2" I consider using of social networking sites at the workplace as a distraction" are:
Correlation Coefficient Sig (P value) Sample Size
-.245** 0.002 127
I am interpreting it as: "there is a moderate negative statistical significant association between statement 1(Variable) and statement 2 (Variable). ( τb= -
245, p
Used to study the relationship and the strength of the relationship between variables
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS