A gentle tutorial on accelerated parameter and confidence interval estimation for hidden Markov models using Template...

Area of Science:

• Computational Statistics
• Statistical Modeling
• Bioinformatics

Background:

• Maximum Likelihood (ML) estimation is a common method for Hidden Markov Models (HMMs).
• Standard R implementations can be computationally intensive, especially for long sequences.
• Deriving confidence intervals often requires computationally expensive bootstrap methods.

Purpose of the Study:

• To demonstrate how to accelerate ML estimation for HMMs using the R package TMB.
• To show how TMB facilitates the simultaneous retrieval of standard errors.
• To provide accessible scripts for users to leverage TMB's computational benefits.

Main Methods:

• Utilizing the TMB (Template Model Builder) package within the R environment.
• Applying TMB for efficient computation of ML estimates for HMM parameters.
• Comparing computational performance with standard R-based ML estimation.

Main Results:

• Significant speed-up in ML estimation computations for HMMs was achieved using TMB.
• TMB allows for straightforward calculation of standard errors alongside parameter estimates.
• Performance was validated on datasets ranging from small (87-240 data points) to large (2000-5000 data points).

Conclusions:

• The TMB package offers a computationally efficient alternative for HMM parameter estimation in R.
• TMB simplifies the process of obtaining standard errors, improving workflow.
• This approach is beneficial for researchers dealing with large HMM datasets.