Publication Date: 2023/06/04
Abstract: Handling simple binary response data with logistic regression has solved many problems encountered in data analysis across various walks of life. However, dealing with ordinal responses, especially when they are more than two levels has remained a big challenge to researchers. This paper therefore focuses attention on the application of cumulative logit response function with proportional odds in order to show its robustness over chi-square, t-tests, percentages and so on, in analyzing likert scale data which is common among users of statistics in social sciences, environmental, and medical sciences. To implement this, 500 random observations on five socio-demographic variables were simulated.In order to justify the use of proportional odds, score test was carried out on the data, and the assumption was not rejected at 5% level (p-value = 0.4222), this justifies the use of the method. Also, the Deviance and Pearson goodness of fit statisticsshow pvalue = 0.7326 and 0.8130 respectively, this reveals that the model fits the data adequately. Moreover, the proportional odds model is fitted and reliable predictions are made. The method is robust for analyzing ordinal response data such as likert scale.
Keywords: Ordinal data, likert scale, proportional odds model, logistic regression.
DOI: https://doi.org/10.5281/zenodo.8058486
PDF: https://ijirst.demo4.arinfotech.co/assets/upload/files/IJISRT23MAY2243.pdf
REFERENCES