---
title: "6 Continuous-time HMMs"
author: "Jan-Ole Fischer"
output: rmarkdown::html_vignette
# output: pdf_document
vignette: >
  %\VignetteIndexEntry{Continuous-time_HMMs}
  %\VignetteEngine{knitr::rmarkdown}
  %\VignetteEncoding{UTF-8}
bibliography: refs.bib
link-citations: yes
---

```{r, include = FALSE}
knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>",
  # fig.path = "img/",
  fig.align = "center",
  fig.dim = c(8, 6),
  out.width = "85%"
)
```

<div style="background-color:#fff3cd; border-left:5px solid #ffc107; padding:10px 15px; margin:15px 0;">
This vignette has not yet been updated to work with RTMB
</div>

> Before diving into this vignette, we recommend reading the vignette [**Introduction to LaMa**](https://janolefi.github.io/LaMa/articles/Intro_to_LaMa.html).

The regular HMM formulation needs a key assumption to be applicable, namely the data need to be observed at regular, equidistant time-points such that the transition probabilities can be interpreted meaningfully w.r.t. a specific time unit. If this is not the case, the model used should account for this by building on a mathematical formulation in continuous time. The obvious choice here is to retain most of the HMM model formulation, but replace the unobserved **discrete-time** Markov chain with a **continuous-time** Markov chain. However, here it is important to note that the so-called **snapshot property** needs to be fulfilled, i.e. the observed process at time $t$ can only depend on the state at that time instant and not on the interval since the previous observation. For more details see @glennie2023hidden.

A continuous-time Markov chain is characterised by a so-called **(infinitesimal) generator matrix** 
$$
Q = \begin{pmatrix}
  q_{11} & q_{12} & \cdots & q_{1N} \\
  q_{21} & q_{22} & \cdots & q_{2N} \\
  \vdots & \vdots & \ddots & \vdots \\
  q_{N1} & q_{N2} & \cdots & q_{NN} \\
\end{pmatrix},
$$
where the diagonal entries are $q_{ii} = - \sum_{j \neq i} q_{ij}$, $q_{ij} \geq 0$ for $i \neq j$. This matrix can be interpreted as the derivative of the transition probability matrix and completely describes the dynamics of the state process. The time-spent in a state $i$ is exponentially distributed with rate $-q_{ii}$ and conditional on leaving the state, the probability to transition to a state $j \neq i$ is $\omega_{ij} = q_{ij} / -q_{ii}$. For a more detailed introduction see @dobrow2016introduction (pp. 265 f.). For observation times $t_1$ and $t_2$, we can then obtain the transition probability matrix between these points via the identity
$$
\Gamma(t_1, t_2) = \exp(Q (t_2 - t_1)),
$$
where $\exp()$ is the matrix exponential. This follows from the so-called **Kolmogorov forward equations**, but for more details see @dobrow2016introduction.

## Example 1: two states

### Setting parameters for simulation

```{r, setup}
# loading the package
library(LaMa)
```

We start by setting parameters to simulate data. In this example, state 1 has a smaller rate and the state dwell time in state one follows and $Exp(0.5)$ distribution, i.e. it exhibits longer dwell times than state 2 with rate 1.
```{r parameters}
# generator matrix Q:
Q = matrix(c(-0.5, 0.5, 1, -1), 
           nrow = 2, byrow = TRUE)

# parameters for the state-dependent (normal) distributions
mu = c(5, 20)
sigma = c(2, 5)
```

### Simulating data

We simulate the continuous-time Markov chain by drawing the exponentially distributed state dwell-times. Within a stay, we can assume whatever structure we like for the observation times, as these are not explicitly modeled. Here we choose to generate them by a Poisson process with rate $\lambda=1$, but this choice is arbitrary. For more details on Poisson point processes, see the MM(M)PP vignette.

```{r data}
set.seed(123)

k = 200 # number of state switches
trans_times = s = rep(NA, k) # time points where the chain transitions
s[1] = sample(1:2, 1) # initial distribution c(0.5, 0.5)
# exponentially distributed waiting times
trans_times[1] = rexp(1, -Q[s[1],s[1]])
n_arrivals = rpois(1, trans_times[1])
obs_times = sort(runif(n_arrivals, 0, trans_times[1]))
x = rnorm(n_arrivals, mu[s[1]], sigma[s[1]])
for(t in 2:k){
  s[t] = c(1,2)[-s[t-1]] # for 2-states, always a state switch when transitioning
  # exponentially distributed waiting times
  trans_times[t] = trans_times[t-1] + rexp(1, -Q[s[t], s[t]])
  n_arrivals = rpois(1, trans_times[t]-trans_times[t-1])
  obs_times = c(obs_times, 
                sort(runif(n_arrivals, trans_times[t-1], trans_times[t])))
  x = c(x, rnorm(n_arrivals, mu[s[t]], sigma[s[t]]))
}
```

Let's visualise the simulated continuous-time HMM:
```{r, vis_ctHMM}
color = c("orange", "deepskyblue")

n = length(obs_times)
plot(obs_times[1:50], x[1:50], pch = 16, bty = "n", xlab = "observation times", 
     ylab = "x", ylim = c(-5,25))
segments(x0 = c(0,trans_times[1:48]), x1 = trans_times[1:49], 
         y0 = rep(-5,50), y1 = rep(-5,50), col = color[s[1:49]], lwd = 4)
legend("topright", lwd = 2, col = color, 
       legend = c("state 1", "state 2"), box.lwd = 0)
```

### Writing the negative log-likelihood function

The likelihood of a continuous-time HMM for observations $x_{t_1}, \dots, x_{t_T}$ at irregular time points $t_1, \dots, t_T$ has the exact same structure as the regular HMM likelihood:
$$
L(\theta) = \delta^{(1)} \Gamma(t_1, t_2) P(x_{t_2}) \Gamma(t_2, t_3) P(x_{t_3}) \dots \Gamma(t_{T-1}, t_T) P(x_{t_T}) 1^t,
$$
where $\delta^{(1)}$, $P$ and $1^t$ are as usual and $\Gamma(t_k, t_{k+1})$ is computed as explained above. Thus we can fit such models using the standard implementation of the general forward algorithm `forward_g()` with time-varying transition probability matrices. We can use the `generator()` function to compute the infinitesimal generator matrix from an unconstrained parameter vector and `tpm_cont()` to compute all matrix exponentials.

```{r, mllk}
nll = function(par, timediff, x, N){
  mu = par[1:N]
  sigma = exp(par[N+1:N])
  Q = generator(par[2*N+1:(N*(N-1))]) # generator matrix
  Pi = stationary_cont(Q) # stationary dist of CT Markov chain
  Qube = tpm_cont(Q, timediff) # this computes exp(Q*timediff)
  allprobs = matrix(1, nrow = length(x), ncol = N)
  ind = which(!is.na(x))
  for(j in 1:N){
    allprobs[ind,j] = dnorm(x[ind], mu[j], sigma[j])
  }
  -forward_g(Pi, Qube, allprobs)
}
```

### Fitting a continuous-time HMM to the data
```{r, model, warning=FALSE}
par = c(mu = c(5, 15), # state-dependent means
        logsigma = c(log(3), log(5)), # state-dependent sds
        qs = c(log(1), log(0.5))) # off-diagonals of Q

timediff = diff(obs_times)

system.time(
  mod <- nlm(nll, par, timediff = timediff, x = x, N = 2)
)
```

### Results

```{r, results}
N = 2
# mu
round(mod$estimate[1:N],2)
# sigma
round(exp(mod$estimate[N+1:N]))
Q = generator(mod$estimate[2*N+1:(N*(N-1))])
round(Q,3)
```

## Example 2: three states

### Setting parameters for simulation

```{r parameters2}
# generator matrix Q:
Q = matrix(c(-0.5,0.2,0.3,
             1,-2, 1,
             0.4, 0.6, -1), nrow = 3, byrow = TRUE)

# parameters for the state-dependent (normal) distributions
mu = c(5, 15, 30)
sigma = c(2, 3, 5)
```

### Simulating data

The simulation is very similar but we now also have to draw which state to transition to, as explained in the beginning.

```{r data2}
set.seed(123)

k = 200 # number of state switches
trans_times = s = rep(NA, k) # time points where the chain transitions
s[1] = sample(1:3, 1) # uniform initial distribution
# exponentially distributed waiting times
trans_times[1] = rexp(1, -Q[s[1],s[1]])
n_arrivals = rpois(1, trans_times[1])
obs_times = sort(runif(n_arrivals, 0, trans_times[1]))
x = rnorm(n_arrivals, mu[s[1]], sigma[s[1]])
for(t in 2:k){
  # off-diagonal elements of the s[t-1] row of Q divided by the diagonal element
  # give the probabilities of the next state
  s[t] = sample(c(1:3)[-s[t-1]], 1, prob = Q[s[t-1],-s[t-1]]/-Q[s[t-1],s[t-1]])
  # exponentially distributed waiting times
  trans_times[t] = trans_times[t-1] + rexp(1, -Q[s[t], s[t]])
  n_arrivals = rpois(1, trans_times[t]-trans_times[t-1])
  obs_times = c(obs_times, 
                sort(runif(n_arrivals, trans_times[t-1], trans_times[t])))
  x = c(x, rnorm(n_arrivals, mu[s[t]], sigma[s[t]]))
}
```

### Fitting a 3-state continuous-time HMM to the data
```{r, model2, warning=FALSE}
par = c(mu = c(5, 10, 25), # state-dependent means
        logsigma = c(log(2), log(2), log(6)), # state-dependent sds
        qs = rep(0, 6)) # off-diagonals of Q

timediff = diff(obs_times)

system.time(
  mod2 <- nlm(nll, par, timediff = timediff, x = x, N = 3, stepmax = 10)
)
# without restricting stepmax, we run into numerical problems
```

### Results

```{r, results2}
N = 3
# mu
round(mod2$estimate[1:N],2)
# sigma
round(exp(mod2$estimate[N+1:N]),2)
Q = generator(mod2$estimate[2*N+1:(N*(N-1))])
round(Q, 3)
```

> Continue reading with [**Markov-modulated Poisson processes**](https://janolefi.github.io/LaMa/articles/MMMPPs.html).

## References