Comments (3)
In theory, you could get some asymptotic standard error estimates from the Hessian used in the numerical optimization (local_step = TRUE). However, this is at least currently not supported and would need some work as the nloptr
used for the numerical optimization does not return the Hessian. Of course, "manually" estimating the model with optim
and logLik
functions with hessian=TRUE
is possible.
If you have a reasonable amount of sequences, you could compute nonparametric bootstrap estimates though.
from seqhmm.
I do have a lot of sequences. In the lowest case, around 50,000. In the largest case, around 1 million. How would I go about doing the bootstrap? I am familiar with bootstrap methods in general, but this is my first trip into HMM methods so I am open to suggestions. My first thought would be to randomly select entire sequences (with replacement), so that the original sequence remains intact but the sample composition becomes random.
Sidenote: The parallelization works well! Estimation is not too slow even with a large number of sequences when using 64 cores. Thanks for that!
from seqhmm.
Yes, your strategy of sampling randomly entire sequences sounds right. In order to avoid potential issues with multiple (local) optima (as well as in order to speed the bootstrap), I suggest you use the estimated parameters as initial values in the bootstrap loop, i.e. you have your estimated model based on the original data, say mod
, and then you define boot_model <- build_hmm(boot_sequences, transition_probs = mod$transition_probs, emission_probs = mod$emission_probs, initial_probs = mod$initial_probs)
.
from seqhmm.
Related Issues (20)
- [Help request] How to test the model? HOT 1
- can this package support Multivariate Discrete HMM?? HOT 3
- will it work for multivariate time series prediction : different continues or/and discrete/category observation HOT 4
- HMM cluster assignments HOT 2
- seqdata should be a state sequence object HOT 1
- Apply to financial time series? HOT 1
- Parsing HMMER3 files HOT 2
- 'System seems singular' and 'EM algorithm failed' for Mixture Markov Model HOT 2
- Building a model with different sequence lengths HOT 1
- hidden_paths does not respect sequence length HOT 4
- Parallel computation HOT 2
- Forward probability of MHMM vs HMM HOT 1
- Maximum number of colours in cpal/colorpalette HOT 4
- Extracting combinations of emitted states HOT 4
- Error in if (em.con$reltol < resEM$change) { : argument is of length zero HOT 5
- Absorbing state broken in `build_mm()` (seqhmm 1.2.1-1) HOT 1
- Runtime Estimation HOT 1
- EM algorithm failed HOT 1
- Error: number of labels must equal number of states in the alphabet HOT 4
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from seqhmm.