I did a linear by linear association test for ordinal data using chi-square in SPSS which was statistically significant. Which Ordinal by ordinal test will be the best test for the measure of strength of association?
Kendall's Tau C or Kendall's Tau B or Gamma and how do we interpret and report the results?
*Data is ordinal and is a 2 by 4 contingency table
Thanks for the replies.
So either
the Gamma, number of concordant pairs minus the discordant pairs, over the total. This would be best in assuming the data's concordancey is a singular finite variable.
The Tau B, number of concordant pairs minus the discordant pairs, over the square root of the multiple of the difference between the max pairs and the number of pairs for each. This would be best in assuming the data's concordancey is are two multiples relative to some overall sum.
The Tau C, number of concordant pairs minus the discordant pairs, over the total number of pairs squared multiplied by the smallest correlation in the set. This would be best in assuming the data's concordancey is scaled by square summable numbers (quantum) and is even better in the case that it's built up out of units common to the lowest unit of concordancey.
So it depends, I'd stick to the standard of Tau C, but if the other options give better results... well... should you even be asking the questions?
Thanks a lot Aden Handasyde !
After posting this question i did some research and went for the Tau C. However i also ran Tau B and still gave the same result. Thanks a lot.
Thanks a lot Matthew Kerry for your response. I will look into that.
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS