If you’re summing multiple Likert items for each variable, you can often treat them as continuous variables. MythBuster’s like (erroneous) “commonsense” conclusions you point out in your book. Your guidance would be most appreciated in helping my group derive actionable insights from our big data set vs. Kruskal-Wallace as the proper method for this space.Īdding to this is I have the constraint that I have Excel (highly skilled) and am a rank beginner in RStudio and I cannot afford SPSS, as I am unaffiliated w/an ed institution. I am beyond the edge of the envelope of my understanding when I read articles about Kendall’s tau vs. I read your book and several other books and on-line resources and I cannot find a clear dummies-friendly “Here is how you should test Categorical x Ordinal non-normal distribution but large data set size” guidance. I want to test if these are likely non-random variance. Simply XLS histograms show variance in independent variable responses distributions to various questions. I want to compare a series of binary independent variables (gender, LGQBT, veteran, etc.) against Likert-like scale attitudinal question/responses (ordinal), and unevenly worded so sometimes they are uni-directional and sometimes bi-directional answer scales.

I have a large (4K response – 32% response rate) survey response population, which is off course not a random sample. I am dusting off my very dusty stats training to help a not-for-profit hiking group analyze membership data. This article and your book(pp 337-341) left off just where my questions begin. In most cases, if there is an actual difference between populations, the two tests have an equal probability of detecting it. Apprehensions about the Mann-Whitney test being underpowered were unsubstantiated. Regarding statistical power, the simulation study shows that there is a minute difference between these two tests. Excessive false positives are not a concern for either hypothesis test. Further, the error rates for both analyses are close to the significance level target. The 2-sample t-test and Mann-Whitney test produce nearly equal false positive rates for Likert scale data. This error rate should equal the significance level. The test results are statistically significant but unbeknownst to the investigator, the null hypothesis is actually true. A type I error rate is essentially a false positive. Comparing Error Rates and Power When Analyzing Likert Scale DataĪfter analyzing all pairs of distributions, the results indicate that both types of analyses produce type I error rates that are nearly equal to the target value.

Likert scale hotel