Mice are commonly used to study infection, as they allow host and pathogen parameters to be evaluated together and over disease progression. Statistical analysis is integral to mouse infection studies and typically endeavours to explain infection phenomena from explanatory variables. A regression model is often employed, which is essentially an optimisation procedure (mathematical part) with the incorporation of uncertainty (statistical part). To obtain robust estimates of the latter, assumptions, such as normality, linearity and homoscedasity in simple linear regression, apply. Although a simple regression model is widely applied in microbiology data, it is not often appropriate or optimal. Here we describe the process of arriving at the most appropriate regression model, using as a case study longitudinal urinalysis data (viable bacterial counts in urine) from mice with experimental urinary tract infection induced by uropathogenic Escherichia coli (UPEC). Groups of wild-type C57BL/6 (n=26) and rag1-/- mice (n = 11) defective in adaptive immune responses, were inoculated with UPEC in the bladder and colony forming units (CFU) in urine were measured over 28 days. Data were analysed using a series of statistical models for parsimony and appropriateness, ranging from a simple regression to mixed models, additive mixed models, and zero-inflated negative binomial mixed models (ZINBMM). The best fitted model was the ZINBMM, which provided an elegant way to answer to two questions: (i) “what is the probability that each strain of mice will be colonised?” and (ii) “once colonised, is there a difference between the two mouse strains with respect to bacterial burden?”. The ZINBMM uncovered a difference in the colonisation probability between mouse strains, but once infection was established, both maintained similar urinalysis profiles. Our findings support a role for adaptive immunity in urinary tract infection control in mice and demonstrate that finding the best-fit statistical model can meaningfully explain biological processes.