fit_hbds_model_parametric {castor} | R Documentation |

Given a timetree (potentially sampled through time and not necessarily ultrametric), fit a homogenous birth-death-sampling (HBDS) model in which speciation, extinction and lineage sampling occurs at some continuous (Poissonian) rates *λ*, *μ* and *ψ*, which are given as parameterized functions of time before present. Sampled lineages are kept in the pool of extant lineages at some “retention probability” *κ*, which may also depend on time. In addition, this model can include concentrated sampling attempts (CSAs) at a finite set of discrete time points *t_1,..,t_m*. “Homogenous” refers to the assumption that, at any given moment in time, all lineages exhibit the same speciation/extinction/sampling rates. Every HBDS model is thus defined based on the values that *λ*, *μ*, *ψ* and *κ* take over time, as well as the sampling probabilities *ρ_1,..,ρ_m* and retention probabilities *κ_1,..,κ_m* during the concentrated sampling attempts; each of these parameters, in turn, is assumed to be determined by a finite set of parameters. This function estimates these parameters by maximizing the corresponding likelihood of the timetree. Special cases of this model are sometimes known as “birth-death-skyline plots” in the literature (Stadler 2013). In epidemiology, these models are often used to describe the phylogenies of viral strains sampled over the course of the epidemic.

fit_hbds_model_parametric(tree, param_values, param_guess = NULL, param_min = -Inf, param_max = +Inf, param_scale = NULL, root_age = NULL, oldest_age = NULL, lambda = 0, mu = 0, psi = 0, kappa = 0, age_grid = NULL, CSA_ages = NULL, CSA_probs = NULL, CSA_kappas = 0, condition = "auto", ODE_relative_dt = 0.001, ODE_relative_dy = 1e-3, CSA_age_epsilon = NULL, Ntrials = 1, max_start_attempts = 1, Nthreads = 1, max_model_runtime = NULL, Nbootstraps = 0, Ntrials_per_bootstrap = NULL, fit_control = list(), focal_param_values = NULL, verbose = FALSE, diagnostics = FALSE, verbose_prefix = "")

`tree` |
A timetree of class "phylo", representing the time-calibrated reconstructed phylogeny of a set of extant and/or extinct species. Tips of the tree are interpreted as terminally sampled lineages, while monofurcating nodes are interpreted as non-terminally sampled lineages, i.e., lineages sampled at some past time point and with subsequently sampled descendants. |

`param_values` |
Numeric vector, specifying fixed values for a some or all model parameters. For fitted (i.e., non-fixed) parameters, use |

`param_guess` |
Numeric vector of size NP, specifying a first guess for the value of each model parameter. For fixed parameters, guess values are ignored. Can be |

`param_min` |
Optional numeric vector of size NP, specifying lower bounds for model parameters. If of size 1, the same lower bound is applied to all parameters. Use |

`param_max` |
Optional numeric vector of size NP, specifying upper bounds for model parameters. If of size 1, the same upper bound is applied to all parameters. Use |

`param_scale` |
Optional numeric vector of size NP, specifying typical scales for model parameters. If of size 1, the same scale is assumed for all parameters. If |

`root_age` |
Positive numeric, specifying the age of the tree's root. Can be used to define a time offset, e.g. if the last tip was not actually sampled at the present. If |

`oldest_age` |
Strictly positive numeric, specifying the oldest time before present (“age”) to consider when calculating the likelihood. If this is equal to or greater than the root age, then |

`lambda` |
Function specifying the speciation rate at any given age (time before present) and for any given parameter values. This function must take exactly two arguments, the 1st one being a numeric vector (one or more ages) and the 2nd one being a numeric vector of size NP (parameter values), and return a numeric vector of the same size as the 1st argument with strictly positive entries. Can also be a single numeric (i.e., lambda is fixed). |

`mu` |
Function specifying the extinction rate at any given age and for any given parameter values. This function must take exactly two arguments, the 1st one being a numeric vector (one or more ages) and the 2nd one being a numeric vector of size NP (parameter values), and return a numeric vector of the same size as the 1st argument with non-negative entries. Can also be a single numeric (i.e., mu is fixed). |

`psi` |
Function specifying the continuous (Poissonian) lineage sampling rate at any given age and for any given parameter values. This function must take exactly two arguments, the 1st one being a numeric vector (one or more ages) and the 2nd one being a numeric vector of size NP (parameter values), and return a numeric vector of the same size as the 1st argument with non-negative entries. Can also be a single numeric (i.e., psi is fixed). |

`kappa` |
Function specifying the retention probability for continuously sampled lineages, at any given age and for any given parameter values. This function must take exactly two arguments, the 1st one being a numeric vector (one or more ages) and the 2nd one being a numeric vector of size NP (parameter values), and return a numeric vector of the same size as the 1st argument with non-negative entries. The retention probability is the probability of a sampled lineage remaining in the pool of extant lineages. Can also be a single numeric (i.e., kappa is fixed). |

`age_grid` |
Numeric vector, specifying ages at which the |

`CSA_ages` |
Optional numeric vector, listing ages (in ascending order) at which concentrated sampling attempts occurred. If |

`CSA_probs` |
Function specifying the sampling probabilities during the various concentrated sampling attempts, depending on parameter values. Hence, for any choice of parameters, |

`CSA_kappas` |
Function specifying the retention probabilities during the various concentrated sampling attempts, depending on parameter values. Hence, for any choice of parameters, |

`condition` |
Character, either "crown", "stem", "none" or "auto", specifying on what to condition the likelihood. If "crown", the likelihood is conditioned on the survival of the two daughter lineages branching off at the root. If "stem", the likelihood is conditioned on the survival of the stem lineage. Note that "crown" really only makes sense when |

`ODE_relative_dt` |
Positive unitless number, specifying the default relative time step for the ordinary differential equation solvers. Typical values are 0.01-0.001. |

`ODE_relative_dy` |
Positive unitless number, specifying the relative difference between subsequent simulated and interpolated values, in internally used ODE solvers. Typical values are 1e-2 to 1e-5. A smaller |

`CSA_age_epsilon` |
Non-negative numeric, in units of time, specfying the age radius around a concentrated sampling attempt, within which to assume that sampling events were due to that concentrated sampling attempt. If |

`Ntrials` |
Integer, specifying the number of independent fitting trials to perform, each starting from a random choice of model parameters. Increasing |

`max_start_attempts` |
Integer, specifying the number of times to attempt finding a valid start point (per trial) before giving up on that trial. Randomly chosen extreme start parameters may occasionally result in Inf/undefined likelihoods, so this option allows the algorithm to keep looking for valid starting points. |

`Nthreads` |
Integer, specifying the number of parallel threads to use for performing multiple fitting trials simultaneously. This should generally not exceed the number of available CPUs on your machine. Parallel computing is not available on the Windows platform. |

`max_model_runtime` |
Optional numeric, specifying the maximum number of seconds to allow for each evaluation of the likelihood function. Use this to abort fitting trials leading to parameter regions where the likelihood takes a long time to evaluate (these are often unlikely parameter regions). |

`Nbootstraps` |
Integer, specifying the number of parametric bootstraps to perform for estimating standard errors and confidence intervals of estimated model parameters. Set to 0 for no bootstrapping. |

`Ntrials_per_bootstrap` |
Integer, specifying the number of fitting trials to perform for each bootstrap sampling. If |

`fit_control` |
Named list containing options for the |

`focal_param_values` |
Optional numeric matrix having NP columns and an arbitrary number of rows, listing combinations of parameter values of particular interest and for which the log-likelihoods should be returned. This may be used for diagnostic purposes, e.g., to examine the shape of the likelihood function. |

`verbose` |
Logical, specifying whether to print progress reports and warnings to the screen. Note that errors always cause a return of the function (see return values |

`diagnostics` |
Logical, specifying whether to print detailed information (such as model likelihoods) at every iteration of the fitting routine. For debugging purposes mainly. |

`verbose_prefix` |
Character, specifying the line prefix for printing progress reports to the screen. |

This function is designed to estimate a finite set of scalar parameters (*p_1,..,p_n\in\R*) that determine the speciation rate *λ*, the extinction rate *μ*, the sampling rate *ψ*, the retention rate *κ*, the concentrated sampling probabilities *ρ_1,..,ρ_m* and the concentrated retention probabilities *κ_1,..,κ_m*, by maximizing the likelihood of observing a given timetree under the HBDS model. Note that the ages (times before present) of the concentrated sampling attempts are assumed to be known and are not fitted.

It is generally advised to provide as much information to the function `fit_hbds_model_parametric`

as possible, including reasonable lower and upper bounds (`param_min`

and `param_max`

), a reasonable parameter guess (`param_guess`

) and reasonable parameter scales `param_scale`

. If some model parameters can vary over multiple orders of magnitude, it is advised to transform them so that they vary across fewer orders of magnitude (e.g., via log-transformation). It is also important that the `age_grid`

is sufficiently fine to capture the variation of *λ*, *μ*, *ψ* and *κ* over time, since the likelihood is calculated under the assumption that these functions vary linearly between grid points.

Note that in this function age always refers to time before present, i.e., present day age is 0 and age increases from tips to root. The functions `lambda`

, `mu`

, `psi`

and `kappa`

should be functions of age, not forward time. Similarly, concentrated sampling attempts (CSAs) are enumerated in order of increasing age, i.e., starting with the youngest CSA and moving towards older CSAs.

A list with the following elements:

`success` |
Logical, indicating whether model fitting succeeded. If |

`objective_value` |
The maximized fitting objective. Currently, only maximum-likelihood estimation is implemented, and hence this will always be the maximized log-likelihood. |

`objective_name` |
The name of the objective that was maximized during fitting. Currently, only maximum-likelihood estimation is implemented, and hence this will always be “loglikelihood”. |

`loglikelihood` |
The log-likelihood of the fitted model for the given timetree. |

`param_fitted` |
Numeric vector of size NP (number of model parameters), listing all fitted or fixed model parameters in their standard order (see details above). If |

`param_guess` |
Numeric vector of size NP, listing guessed or fixed values for all model parameters in their standard order. If |

`guess_loglikelihood` |
The loglikelihood of the data for the initial parameter guess ( |

`focal_loglikelihoods` |
A numeric vector of the same size as |

`NFP` |
Integer, number of fitted (i.e., non-fixed) model parameters. |

`Ndata` |
Number of data points used for fitting, i.e., the number of sampling and branching events that occurred between ages 0 and |

`AIC` |
The Akaike Information Criterion for the fitted model, defined as |

`BIC` |
The Bayesian information criterion for the fitted model, defined as |

`condition` |
Character, specifying what conditioning was root for the likelihood (e.g. "crown" or "stem"). |

`converged` |
Logical, specifying whether the maximum likelihood was reached after convergence of the optimization algorithm. Note that in some cases the maximum likelihood may have been achieved by an optimization path that did not yet converge (in which case it's advisable to increase |

`Niterations` |
Integer, specifying the number of iterations performed during the optimization path that yielded the maximum likelihood. |

`Nevaluations` |
Integer, specifying the number of likelihood evaluations performed during the optimization path that yielded the maximum likelihood. |

`trial_start_objectives` |
Numeric vector of size |

`trial_objective_values` |
Numeric vector of size |

`trial_Nstart_attempts` |
Integer vector of size |

`trial_Niterations` |
Integer vector of size |

`trial_Nevaluations` |
Integer vector of size |

`standard_errors` |
Numeric vector of size NP, estimated standard error of the parameters, based on parametric bootstrapping. Only returned if |

`medians` |
Numeric vector of size NP, median the estimated parameters across parametric bootstraps. Only returned if |

`CI50lower` |
Numeric vector of size NP, lower bound of the 50% confidence interval (25-75% percentile) for the parameters, based on parametric bootstrapping. Only returned if |

`CI50upper` |
Numeric vector of size NP, upper bound of the 50% confidence interval for the parameters, based on parametric bootstrapping. Only returned if |

`CI95lower` |
Numeric vector of size NP, lower bound of the 95% confidence interval (2.5-97.5% percentile) for the parameters, based on parametric bootstrapping. Only returned if |

`CI95upper` |
Numeric vector of size NP, upper bound of the 95% confidence interval for the parameters, based on parametric bootstrapping. Only returned if |

`consistency` |
Numeric between 0 and 1, estimated consistency of the data with the fitted model (Lindholm et al. 2019). See the documentation of |

Stilianos Louca

T. Stadler, D. Kuehnert, S. Bonhoeffer, A. J. Drummond (2013). Birth-death skyline plot reveals temporal changes of epidemic spread in HIV and hepatitis C virus (HCV). PNAS. 110:228-233.

A. Lindholm, D. Zachariah, P. Stoica, T. B. Schoen (2019). Data consistency approach to model validation. IEEE Access. 7:59788-59796.

`generate_tree_hbds`

,
`simulate_deterministic_hbds`

## Not run: # Generate a random tree with exponentially varying lambda & mu and constant psi # assume that all sampled lineages are removed from the pool (i.e. kappa=0) time_grid = seq(from=0, to=100, by=0.01) root_age = 5 tree = generate_tree_hbds(max_time = root_age, time_grid = time_grid, lambda = 2*exp(0.1*time_grid), mu = 0.1*exp(0.09*time_grid), psi = 0.1, kappa = 0)$tree cat(sprintf("Tree has %d tips\n",length(tree$tip.label))) # Define a parametric HBDS model, with exponentially varying lambda & mu # Assume that the sampling rate is constant but unknown # The model thus has 5 parameters: lambda0, mu0, alpha, beta, psi lambda_function = function(ages,params){ return(params['lambda0']*exp(-params['alpha']*ages)); } mu_function = function(ages,params){ return(params['mu0']*exp(-params['beta']*ages)); } psi_function = function(ages,params){ return(rep(params['psi'],length(ages))) } # Define an age grid on which lambda_function & mu_function shall be evaluated # Should be sufficiently fine to capture the variation in lambda & mu age_grid = seq(from=0,to=root_age,by=0.01) # Perform fitting cat(sprintf("Fitting model to tree..\n")) fit = fit_hbds_model_parametric(tree, root_age = root_age, param_values = c(lambda0=NA, mu0=NA, alpha=NA, beta=NA, psi=NA), param_guess = c(1,1,0,0,0.5), param_min = c(0,0,-1,-1,0), param_max = c(10,10,1,1,10), param_scale = 1, # all params are in the order of 1 lambda = lambda_function, mu = mu_function, psi = psi_function, kappa = 0, age_grid = age_grid, Ntrials = 4, # perform 4 fitting trials Nthreads = 2) # use 2 CPUs if(!fit$success){ cat(sprintf("ERROR: Fitting failed: %s\n",fit$error)) }else{ cat(sprintf("Fitting succeeded:\nLoglikelihood=%g\n",fit$loglikelihood)) # print fitted parameters print(fit$param_fitted) } ## End(Not run)

[Package *castor* version 1.6.8 Index]