Permutation test in evaluating the significance of plants in pls-da model Of jamu ingredients

Abstract

PLS-DA (Partial Least Square Discriminant Analysis) model is used to establish relationship between Jamu ingredients, i.e. plants composition in Jamu, and Jamu efficacy to investigate which plants act as main ingredients and which are supporting by checking the plants significance. Permutation testing is used in the investigation by generating the coefficients distribution under null hypothesis, i.e. the plants are not affecting Jamu efficacy. The generation process is performed by permuting the order of the response while maintaining the order of the predictors. The PLS-DA model then is applied to the new dataset after permutation. After repeating this process many times, then the accumulation of the PLS-DA coefficients provides the distribution under the null hypothesis. The proportion of the coefficients larger than or equal to the PLS-DA coefficient using original data then serves as the p-value, which then can be compared to the significance level . By performing this permutation process 1000 times and  = 5%, we found, over all efficacies, 231 out of 465 plants are significant. Moreover, from literature review, among these 231 plants, the usages of 226 plants on the assigned efficacy are supported by scientific paper.