i have a data of customer responses measured on likert scale with respect to some service of the bank. customers are divided into two groups viz., current account customers and saving bank account customers.
sample of current account customers is 140 and that of saving bank account is 260.
I want to test whether there is any significant difference in the response of the customers of the two groups with respect to each of the banking service
thanks in advance
The appropriate test depends on what you mean by "Likert scale."
If you are looking at a single Likert item, or maybe a few that are combined, the dependent variable should be treated as ordinal variable.
In this case you can arrange counts in to a contingency table and use the Cochran-Armitage test. For a measure of effect size, you can Freeman's theta or epsilon-squared.
In this case, you could also use Mann-Whitney test, although some people object to using this test for dependent variables where there are few different levels in the dependent variable.
If you have created a scale by combining several Likert items, and there are many different levels in the answers, you can treat the dependent variable as a continuous variable. In this case, Mann-Whitney would be an appropriate test.
Because your sample size is relatively large, you are likely to get a significant p-value even with a small difference between groups. Because of this, you should be sure to report an effect size or the the difference between the mean or median between to the two groups.
T-Test is your appropriate test for comparing differences in the means of two groups.
T-test examines whether the population means of two samples greatly differ from one another, using t-distribution which is used when the standard deviation is not known, and the sample size is small. It is a tool to analyse whether the two samples are drawn from the same population.
The test is based on t-statistic, which assumes that variable is normally distributed (symmetric bell-shaped distribution) and mean is known and population variance is calculated from the sample.
References (APA Style)
1. Ott, R. L., & Longnecker, M. T. (2015). An introduction to statistical methods and data analysis. Nelson Education. Book page 340
2. Vogt, W. P., Gardner, D. C., Haeffele, L. M., & Vogt, E. R. (2014). Selecting the right analyses for your data: Quantitative, qualitative, and mixed methods. Guilford Publications. Book page 275
I think the way to go is for you to generate new variable from the question items. Your selection of appropriate statistical method here depends on whether you want to treat the newly generated variable as continuous which of course it is or you want to dichotomize it assuming you have information that will help you that meaningfully.
If your new variable is treated as a continuous outcome variable then t test could be appropriate of course normality assumption must be established. If for some reasons you dichotomized then chi square method would be appropriate. If on the other hand you have some other information on your customers apart from their type of account and such information are believed to have some predictive tendencies on the outcome, then, you may want to consider a multivariate technique; linear regression for continuous outcome variable or binary logistic for a binary outcome variable. I hope this helps.
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Modeling
- Analyses of Variance
- ANOVA