merlin—A unified modeling framework for data analysis and methods development in Stata

Abstract

The challenges in statistics and data science are rapidly growing because access to a multitude of data types continues to increase, as well as the sheer quantity of data. Analysts are now presented with multivariate data, sometimes measured repeatedly, and often requiring the ability to model nonlinear relationships and hierarchical structures. In this article, I present the merlin command, which attempts to provide an extremely general framework for data analysis. From simple settings such as fitting a linear regression model or a Weibull survival model to more complex settings such as fitting a three-level logistic mixed-effects model or a multivariate joint model of multiple longitudinal outcomes (of different types) and a recurrent event and survival with nonlinear effects, merlin can fit them all. I will take a single dataset and attempt to show you the full range of capabilities of merlin and discuss some future directions for the implementation in Stata.

Assessing relative COVID-19 mortality during the second wave: a prospective Swiss population-based study

“… 30 34 Multiplicity-adjusted p values, CIs and confidence bands were computed using the multcomp package. 35 Analyses were independently replicated in Stata 36 (V.16), using stpm2 37 38 and merlin 39 .…”

“…30 34 Multiplicity-adjusted p values, CIs and confidence bands were computed using the multcomp package. 35 Analyses were independently replicated in Stata 36 (V.16), using stpm2 37 38 and merlin 39 . The age distribution of men in the Swiss Fall 2020 COVID-19 cohort matched the age distribution in the Swiss Fall 2018 cohort rather closely (figure 2, right panel), whereas women aged 65-75 years were slightly under-represented in the COVID-19 cohort (figure 2, left panel).…”

Effect of viral replication and liver fibrosis on all-cause mortality in HIV/HBV coinfected patients: a retrospective analysis of a 15-year longitudinal cohort

“…We included a random-intercept across all models to account for between-patient variance at baseline, while the random-intercepts were constrained at 1 for the two submodels. Parameters from the three models were jointly estimated via maximum likelihood using the ‘merlin’ program in STATA [21].…”

Joint modelling of multivariate longitudinal clinical laboratory safety outcomes, concomitant medication and clinical adverse events: application to artemisinin-based treatment during pregnancy clinical trial

“…To mitigate this problem, we employed Monte Carlo integration technique since the number of draws it makes from the random effects do not need to change with increase in number of random effects [ 23 , 24 ] (hence reducing the computation burden). Estimation of our models was done using merlin in Stata [ 21 ]. Only 7 patients had missing data points and the data missing mechanism was considered completely at random.…”

“…Despite the advanced development and evident benefits of joint modelling of multivariate longitudinal data [ 17 – 19 ], joint modelling has received limited attention in drug safety assessment. Recent statistical software developments have provided an opportunity for extending joint models to more different types of outcomes [ 20 , 21 ], beyond the traditionally implemented joint models with single continuous longitudinal data and single survival outcome. In this paper, we present a joint model for longitudinal continuous clinical laboratory safety data (alanine amino transferase and total bilirubin), individual clinical AEs and concomitant medication.…”