Determination Of The Facial Foam Sensory Attributes’ Optimum Value Using Ordinal Logistic Regression Models
Ekaputri, Andini Desita
MetadataShow full item record
ANDINI DESITA EKAPUTRI. Determination of the Facial Foam Sensory Attributes Optimum Value Using Ordinal Logistic Regression Models. Under the supervision of KHAIRIL ANWAR NOTODIPUTRO and BAMBANG SUMANTRI. Market products have many characteristics to be maintained by the companies to fit the consumers’ preference. One of important characteristic s is sensory aspects. Sensory aspects are built by the five senses from the human body namely sight, sound, taste, touch and smell. However, the problem is the optimum value of each sensory attribute is still unknown. Preference attribute, which is the response variable is usually measured by an ordinal scale. As a result simple linear regression would not be appropriate. One of the alternatives is by using ordinal logistic regression, a model that takes rank ordering of the response variables into account. Using one of the ordinal logistic models, the proportional odds model, the optimum value will be obtained by searching the highest individual probability for the “like it” category of preference attribute for each category of performance attribute. The analysis for each of the attribute were treated differently depended on the brand product influence, named panel. If an attribute was influenced by the panel then the optimum value would be obtained for each panel. However, if all panels did not give any significant influences, than the analysis could be done entirely without considering the panels. The result showed that only three sensory attributes were affected by the panels, namely smoothness, thickness of foam during use and ease of rinsing at tributes. The optimum value of a product was based on the individual probability of the preference attribute. In other word the objective was to know how big was the probability to choose scale 4 of the preference attribute, which was the “like it category”, of each performance attribute’s category. For most of the performance attributes, scale 4 tended to have largest probability to be preferred. Only stickiness and sticky feel after rinsing had the largest probability on scale 1 which was “not sticky at all category ”. The last objectives was to find out the association among categories of performance attribute using odds ratio. An odds ratio that has value more than one means that the observed category is less preferred than the reference category. C onversely, an odds ratio that has value less than one means the observed category is more preferred than the reference category.