Table II.

Factors involved in the choice of statistical model.

Type of analysis	Data type	Advantages	Disadvantages
Mixed effects models	Continuous (linear mixed effects model), categorical or count (non-linear mixed effects models)	•   Allows for imbalanced data size and informative missingness (i.e. the number of cycles a woman contributes depends on the outcomes of her prior cycles) •   Models can specify multiple correlations	•   More difficult to implement and require sufficient statistical training •   Models are sometimes unstable and may not converge
Generalized estimating equations (GEEs)	Continuous, categorical, count	•   Can be used as an alternative to non-linear mixed models when they do not converge •   Have a different interpretation than mixed effects models (population-averaged versus individual-level effects)
Discrete survival	Time-to-event (e.g. first live birth)	•   Easy to implement •   Censors on the outcome, making it appropriate for the use of first live birth as an outcome	•   Can only accommodate one event at a time, though many ART events can be experienced in multiple cycles, such as implantation failure or live birth