| 1 | MrBayes |
|---|
| 2 | |
|---|
| 3 | DESCRIPTION |
|---|
| 4 | MrBayes is a program for Bayesian inference and model choice across |
|---|
| 5 | a wide range of phylogenetic and evolutionary models. MrBayes uses |
|---|
| 6 | Markov chain Monte Carlo (MCMC) methods to estimate the posterior |
|---|
| 7 | distribution of model parameters. |
|---|
| 8 | |
|---|
| 9 | The desciptions in this manual where copied from the official |
|---|
| 10 | MrBayes manual at LINK{http://mrbayes.sourceforge.net/manual.php} |
|---|
| 11 | |
|---|
| 12 | |
|---|
| 13 | PARAMETERS |
|---|
| 14 | |
|---|
| 15 | Number of substitution types |
|---|
| 16 | |
|---|
| 17 | Sets the number of substitution types: |
|---|
| 18 | |
|---|
| 19 | - "1" constrains all of the rates to be the same (e.g., a JC69 or F81 model); |
|---|
| 20 | - "2" allows transitions and transversions to have potentially different rates |
|---|
| 21 | (e.g., a K80 or HKY85 model); |
|---|
| 22 | - "6" allows all rates to be different, subject to the constraint of |
|---|
| 23 | time-reversibility (e.g., a GTR model). |
|---|
| 24 | - Finally, 'nst' can be set to 'mixed', which results in the Markov |
|---|
| 25 | chain sampling over the space of all possible reversible |
|---|
| 26 | substitution models, including the GTR model and all models that |
|---|
| 27 | can be derived from it model by grouping the six rates in various |
|---|
| 28 | combinations. This includes all the named models above and a large |
|---|
| 29 | number of others, with or without name. |
|---|
| 30 | |
|---|
| 31 | Model for among-site rate variation |
|---|
| 32 | |
|---|
| 33 | Sets the model for among-site rate variation. |
|---|
| 34 | In general, the rate at a site is considered to be an unknown random |
|---|
| 35 | variable. The valid options are: |
|---|
| 36 | * No rate variation across sites. |
|---|
| 37 | * Gamma-distributed rates across sites. The rate |
|---|
| 38 | at a site is drawn from a gamma distribution. |
|---|
| 39 | The gamma distribution has a single parameter that |
|---|
| 40 | describes how much rates vary. |
|---|
| 41 | * Autocorrelated rates across sites. The marginal rate distribution |
|---|
| 42 | is gamma, but adjacent sites have correlated rates. |
|---|
| 43 | * A proportion of the sites are invariable. |
|---|
| 44 | * Mixed invariable/gamma: A proportion of the sites are invariable while |
|---|
| 45 | the rates for the remaining sites are drawn from a gamma distribution. |
|---|
| 46 | |
|---|
| 47 | Number of rate categories for the gamma distribution |
|---|
| 48 | |
|---|
| 49 | Sets the number of rate categories for the gamma distribution. |
|---|
| 50 | The gamma distribution is continuous. However, it is virtually |
|---|
| 51 | impossible to calculate likelihoods under the continuous gamma |
|---|
| 52 | distribution. Hence, an approximation to the continuous gamma is used; |
|---|
| 53 | the gamma distribution is broken into ncat categories of equal |
|---|
| 54 | weight (1/ncat). The mean rate for each category represents the |
|---|
| 55 | rate for the entire cateogry. This option allows you to specify |
|---|
| 56 | how many rate categories to use when approximating the gamma. |
|---|
| 57 | The approximation is better as ncat is increased. In practice, |
|---|
| 58 | "ncat=4" does a reasonable job of approximating the continuous gamma. |
|---|
| 59 | |
|---|
| 60 | Number of cycles for the MCMC algorithm |
|---|
| 61 | |
|---|
| 62 | This option sets the number of cycles for the MCMC algorithm. |
|---|
| 63 | This should be a big number as you want the chain to first reach |
|---|
| 64 | stationarity, and then remain there for enough time to take lots of samples. |
|---|
| 65 | |
|---|
| 66 | NOTE: the standalone version of MrBayes asks if you want to continue |
|---|
| 67 | the calculation after the number of cycles has been reached. |
|---|
| 68 | This does NOT happen when using the ARB version. If the number |
|---|
| 69 | of cycles has been reached the algorithm will terminate! |
|---|
| 70 | |
|---|
| 71 | Number of chains |
|---|
| 72 | |
|---|
| 73 | How many chains are run for each analysis for the MCMCMC variant. |
|---|
| 74 | The default is 4: 1 cold chain and 3 heated chains. |
|---|
| 75 | If Nchains is set to 1, MrBayes will use regular MCMC sampling, without heating. |
|---|
| 76 | |
|---|
| 77 | Temperature parameter for heating the chains |
|---|
| 78 | |
|---|
| 79 | The temperature parameter for heating the chains. The higher the |
|---|
| 80 | temperature, the more likely the heated chains are to move between |
|---|
| 81 | isolated peaks in the posterior distribution. However, excessive |
|---|
| 82 | heating may lead to very low acceptance rates for swaps between |
|---|
| 83 | different chains. Before changing the default setting, however, |
|---|
| 84 | note that the acceptance rates of swaps tend to fluctuate during |
|---|
| 85 | the burn-in phase of the run. |
|---|
| 86 | |
|---|
| 87 | Markov chain sample frequency |
|---|
| 88 | |
|---|
| 89 | This specifies how often the Markov chain is sampled. You can |
|---|
| 90 | sample the chain every cycle, but this results in very large |
|---|
| 91 | output files. Thinning the chain is a way of making these files |
|---|
| 92 | smaller and making the samples more independent. |
|---|
| 93 | |
|---|
| 94 | Fraction of samples that will be discarded |
|---|
| 95 | |
|---|
| 96 | Determines the fraction of samples that will be discarded when |
|---|
| 97 | convergence diagnostics are calculated. The value of this option |
|---|
| 98 | is only relevant when Relburnin is set to YES. Example: A value |
|---|
| 99 | for this option of 0.25 means that 25% of the samples will be discarded. |
|---|
| 100 | |
|---|
| 101 | |
|---|
| 102 | |
|---|
| 103 | |
|---|
| 104 | |
|---|
| 105 | LICENSE |
|---|
| 106 | GNU GENERAL PUBLIC LICENSE |
|---|
| 107 | Version 2, June 1991 |
|---|
