Fitting a lognormal distribution to enumeration and absence/presence data.

Area of Science:

• Microbiology
• Statistical Modeling
• Food Safety

Background:

• Accurate microbial data analysis is crucial for food safety and risk assessment.
• Traditional methods for fitting lognormal distributions to microbial data often involve data transformation and handling of censored values.
• Existing models may not fully leverage the information present in raw microbial detection and enumeration data.

Purpose of the Study:

• To compare a novel model (M(RD)) that directly uses raw microbial data with a conventional model (M(CLD)) that uses censored concentration data.
• To evaluate the performance of both models in fitting lognormal distributions to complex microbial datasets.
• To determine which model provides less biased and more precise estimates for microbial quantification.

Main Methods:

• Development and application of two distinct statistical models: M(CLD) using censored concentrations and M(RD) using raw data (presence/absence, colony counts).
• Utilized maximum likelihood estimation for M(CLD) and the Expectation-Maximisation (EM) algorithm for M(RD).
• Simulated complex microbial datasets under defined limits of detection (LOD) and quantification (LOQ) for comparative analysis.

Main Results:

• Model M(RD) provided estimates similar to the M(CLD) model in most scenarios when accounting for data censorship.
• In specific cases, M(RD) demonstrated reduced bias and improved precision compared to M(CLD).
• The direct use of raw data in M(RD) appears to offer advantages in certain complex microbial data analyses.

Conclusions:

• The M(RD) model offers a robust alternative for fitting lognormal distributions to microbial data, directly utilizing raw detection and enumeration information.
• M(RD) can achieve comparable or superior statistical performance (bias, precision) relative to traditional censored data models.
• This advancement has implications for more accurate microbial risk assessments and data interpretation in microbiology and food science.