I have two likert type statement questions (ordinal by ordinal ) and they have the following available responses (Strongly Agree (1), Somewhat Agree(2), Neutral (3), Somewhat Disagree (4) and Strongly Disagree (5). Sample size 127
How should I interpret the Kendall Tau b results while comparing these two likert type statements?
For example _ Kendall Tau b = ( -.495)
I know that it will be interpreted as negative, moderate in strength but I want to know how it can be interpreted from the explanation point of view (meaning what it will tell about the likert type statement). Will it mean that respondents/survey participants gave dissimilar ranking to both the statements? (meaning if the respondents gave scoring to statement 1 in the order 1,2,3,4,5 and then gave opposite scoring to statement 2 (5,4,3,2,1).
Hello Amit,
Yes, pretty much as you propose.
1. Yes, the ratings given to the two statements tended to be dissimilar.
2. Yes, higher ratings on one statement tended to be observed in conjunction with lower ratings on the other (or vice versa). (I'd stay away from "opposite," however.)
Good luck with your work.
I am thankful to you. You help so many of us to take the most informed path in our research. David Morse
I would suggest using crosstabulation and Chi-Square test for independence. The test is going to confirm the dependence of variables, but the standardized residuals will show you exactly what you need: which pairs/combinations of ratings have been observed as common or rare in your sample (in comparison to the hypothetical situation where those variables are independent).
Good luck!
Thank you Aleksandar Tomasevic for your answer. But I needed to share that I am finding it difficult to run the Chi-square test for independence because in most of the cases the expected count comes down to less than 5 (sample size is 127 with 5 response categories). Another reason is that I have run Kendall Tau in SPSS using the correlation matrix (for 23 Likert type statements) and I have got the results in the form of a table for all the statements in one go. The chi-square will require me to run the test for each and every statement with each and every other statement.
You can create a 5 x 5 table to be analyzed with Fisher's exact test, due to the small size of some of the categories. More to the point, however, you should be able to examine the frequencies in the table by themselves to determine if there is a negative direction in the relationship. If you have an independent and dependent variable, then calculating percentages within the dependent variable should be helpful.
David L Morgan Thank you for your reply. However, some scholars have suggested to use Fisher's exact test for 2x2 tables. Is that correct?
the need for Fisher's exact test is based on the number of expected cases per cell, not the size of the table.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Sampling Design
- Sample Size