The context of my data is from an e-commerce website. I am measuring the number of errors made based on different independent variables: Filtering technique (Alphabetical, Relevance and Search) and Number of Items added (Add one at a time, or add multiple items to cart at once). As the number of errors made is 0 most of the time, I have opted to use the Zero Inflated Poisson Regression. When I performed the pairwise Tukey HSD comparison, I get the following picture.
How can I interpret the t ratio and p values in this case? For example, for Filtering Technique = Alphabetical and Number of Items added = Multiple, vs Filtering Technique = Search and Number of Items Added = Multiple, t ratio is -3.28 and p=0.0143. Does this mean that Search,Multiple number of items added has a higher significance of errors than Alphabetical,Multiple number of items added (since t is negative)? This seems to contradict the data I have, where there there is no errors made in the scenario where filtering technique is search and there are multiple number of modules added
turkey hsd pairwise tests
ConstructChart2_0-1681414901894.png
Distribution of data (Alphabetical, Multiple)
ConstructChart2_1-1681415087104.png
Distribution of Data (Search, Multiple). Note there are 0 errors made
ConstructChart2_2-1681415116602.png