/archive/source.archive/bm.and.ou/trait.rjmcmc.09162010.R

http://github.com/eastman/auteur · R · 496 lines · 421 code · 58 blank · 17 comment · 68 complexity · 4327115b691c65e3dd0041bc4e0d9444 MD5 · raw file 

    

  1. ## Written by LJ HARMON & AH: 2008

    2. 

  2. ## Modified by JM EASTMAN: 09.2010

    3. 

  3. 

    4. 

  4. rjMCMC.trait<-function (phy, data, ngen=1000, sampleFreq=100, model=c("OU","BM"), probAlpha=0.05, probMergeSplit=0.05, probRoot=0.1, heat=1, fileBase="result") 

    5. 

  5. {

    6. 

  6. 	if(length(model)>1)stop(paste("Please specify either ", sQuote("OU"), " or ", sQuote("BM")))

    7. 

  7. ## MORE ERROR CHECKING NEEDED HERE ## 

    8. 

  8. 	

    9. 

  9. 	require(ouch)

    10. 

  10. 	require(geiger)

    11. 

  11. 	

    12. 

  12. ### prepare objects for rjMCMC

    13. 

  13. 	cur.model		<- model								# initialize family of models explored

    14. 

  14. 

    15. 

  15. 	dataList		<- prepare.ouchdata(phy, data)			# check tree and data; convert into S4

    16. 

  16. 

    17. 

  17. 	ape.tre			<- cur.tre <- dataList$ape.tre			# S3 phylogeny

    18. 

  18. 	orig.dat		<- dataList$orig.dat					# S3 data	

    19. 

  19. 	ouch.tre		<- dataList$ouch.tre					# S4 OUCH phylogeny

    20. 

  20. 	ouch.dat		<- dataList$ouch.dat					# S4 OUCH data

    21. 

  21. 	

    22. 

  22. 	nn				<- length(ape.tre$edge.length)			# max number of rates (one per branch)

    23. 

  23. 	init.rate		<- fitContinuous(ape.tre,orig.dat)

    24. 

  24. 	init.rate		<- adjust.rate(init.rate[[1]]$beta)

    25. 

  25.     cur.rates		<- rep(init.rate,nn)					# initialize array for rate estimates branchwise

    26. 

  26. 	cur.delta.rates	<- rep(0,nn)

    27. 

  27. 	cur.root		<- mean(orig.dat)

    28. 

  28. 	cur.alpha		<- 0.001								# initialize alpha

    29. 

  29. 	cur.theta		<- rep(0,nn)							# initialize trait optima under OU

    30. 

  30. 	cur.delta.theta	<- rep(0,nn)							# initialize local 'rates'

    31. 

  31. 	cur.regimes		<- as.factor(c(sample(cur.theta,1),cur.theta))				# initial set of selective regimes

    32. 

  32. 	cur.root.theta	<- sample(cur.theta,1)

    33. 

  33. 

    34. 

  34. 	nRateProp = 0											# proposal count: update rate parameter(s)

    35. 

  35. 	nRcatProp = 0											# proposal count: update rate categories

    36. 

  36. 	nRootProp = 0

    37. 

  37. 	nAlphaProp = 0											# proposal count: update alpha parameter

    38. 

  38. 	nRegimesProp = 0										# proposal count: update selective regimes

    39. 

  39. 	nThetaProp = 0

    40. 

  40. 	nRateOK = 0												# proposal count: successful rate changes

    41. 

  41. 	nRcatOK = 0												# proposal count: successful rate categories updates

    42. 

  42. 	nRootOK = 0

    43. 

  43. 	nAlphaOK = 0											# proposal count: successful alpha updates

    44. 

  44. 	nRegimesOK = 0											# proposal count: successful regime updates

    45. 

  45. 	nThetaOK = 0

    46. 

  46. 	l <- array(dim=c(ngen,4))								# array for storing output: size is number of MCMC generations

    47. 

  47.     l[] <- NA												# initialize array for storing lnLs

    48. 

  48. 	

    49. 

  49. 	cOK=c(													# list of proposal tallies

    50. 

  50. 		  nRateOK, 

    51. 

  51. 		  nRcatOK, 

    52. 

  52. 		  nRootOK, 

    53. 

  53. 		  nAlphaOK, 

    54. 

  54. 		  nRegimesOK, 

    55. 

  55. 		  nThetaOK

    56. 

  56. 		  )		

    57. 

  57. 	

    58. 

  58. 	errorLog=file(paste(getwd(),paste(cur.model, "rjTraitErrors.log",sep="."),sep="/"),open='w+')

    59. 

  59. 	generate.error.message(i=NULL, mod.cur=NULL, mod.new=NULL, lnR=NULL, errorLog=errorLog, init=TRUE)

    60. 

  60. 	runLog=file(paste(getwd(),paste(cur.model, "rjTrait.log",sep="."),sep="/"),open='w+')

    61. 

  61. 	generate.log(bundled.parms=NULL, cur.model, file=runLog, init=TRUE)

    62. 

  62. 	

    63. 

  63. ### Begin rjMCMC

    64. 

  64.     for (i in 1:ngen) {

    65. 

  65.         lnLikelihoodRatio <- lnHastingsRatio <- lnPriorRatio <- 0

    66. 

  66. 		startparms=list(rates=cur.rates, cats=sum(cur.delta.rates)+1, root=cur.root, alph=cur.alpha, reg=length(table(cur.regimes)), theta=mean(cur.theta))

    67. 

  67. 	

    68. 

  68. ## OU IMPLEMENTATION ## 

    69. 

  69. 		if(cur.model=="OU") {

    70. 

  70. 			if (runif(1) < (2 * probMergeSplit)) {

    71. 

  71. 				if(runif(1)<0.5) {				# adjust theta categories

    72. 

  72. 					nr=split.or.merge(cur.delta.theta, cur.theta, ape.tre, theta=TRUE)

    73. 

  73. 					new.alpha=cur.alpha

    74. 

  74. 					new.root.theta=cur.root.theta

    75. 

  75. 					new.theta=nr$new.values ##

    76. 

  76. 					new.delta.theta=nr$new.delta ##

    77. 

  77. 					nrt=assign.root.theta(new.theta, ape.tre) ## 

    78. 

  78. 					new.regimes=as.factor(c(nrt,new.theta)) ##

    79. 

  79. 					new.rates=cur.rates 

    80. 

  80. 					new.delta.rates=cur.delta.rates 

    81. 

  81. 					nRegimesProp=nRegimesProp+1

    82. 

  82. 				} else {						# adjust rate categories

    83. 

  83. 					nr=split.or.merge(cur.delta.rates, cur.rates, ape.tre, theta=FALSE)

    84. 

  84. 					new.alpha=cur.alpha

    85. 

  85. 					new.root.theta=cur.root.theta

    86. 

  86. 					new.theta=cur.theta

    87. 

  87. 					new.delta.theta=cur.delta.theta

    88. 

  88. 					new.regimes=cur.regimes

    89. 

  89. 					new.rates=nr$new.values ##

    90. 

  90. 					new.delta.rates=nr$new.delta ##

    91. 

  91. 					nRcatProp=nRcatProp+1

    92. 

  92. 				}

    93. 

  93. 				lnHastingsRatio=nr$lnHastingsRatio

    94. 

  94. 				lnPriorRatio=nr$lnPriorRatio

    95. 

  95. 			} else { # neither split nor merge

    96. 

  96. 				if(runif(1)<probAlpha) {		# adjust alpha

    97. 

  97. 					new.alpha=adjust.rate(cur.alpha) ##

    98. 

  98. 					new.root.theta=cur.root.theta

    99. 

  99. 					new.theta=cur.theta

    100. 

  100. 					new.delta.theta=cur.delta.theta

    101. 

  101. 					new.regimes=cur.regimes 

    102. 

  102. 					new.rates=cur.rates

    103. 

  103. 					new.delta.rates=cur.delta.rates

    104. 

  104. 					nAlphaProp=nAlphaProp+1

    105. 

  105. 				} else { 

    106. 

  106. 					if(runif(1)<0.5) {			# adjust rates

    107. 

  107. 						new.alpha=cur.alpha

    108. 

  108. 						new.root.theta=cur.root.theta

    109. 

  109. 						new.theta=cur.theta

    110. 

  110. 						new.delta.theta=cur.delta.theta

    111. 

  111. 						new.regimes=cur.regimes 

    112. 

  112. 						new.rates=adjust.rate(cur.rates) ##

    113. 

  113. 						new.delta.rates=cur.delta.rates

    114. 

  114. 						nRateProp=nRateProp+1

    115. 

  115. 					} else {					# adjust theta

    116. 

  116. 						new.alpha=cur.alpha 

    117. 

  117. 						new.root.theta=cur.root.theta

    118. 

  118. 						new.theta=adjust.value(cur.theta) ##

    119. 

  119. 						new.delta.theta=cur.delta.theta

    120. 

  120. 						nrt=assign.root.theta(new.theta, ape.tre) ## 

    121. 

  121. 						new.regimes=as.factor(c(nrt,new.theta)) ##

    122. 

  122. 						new.rates=cur.rates

    123. 

  123. 						new.delta.rates=cur.delta.rates

    124. 

  124. 						nThetaProp=nThetaProp+1				

    125. 

  125. 					}

    126. 

  126. 				}

    127. 

  127. 			}

    128. 

  128. 			

    129. 

  129. 			cur.tre=apply.BMM.to.apetree(ape.tre, cur.rates)

    130. 

  130. 			new.tre=apply.BMM.to.apetree(ape.tre, new.rates)

    131. 

  131. 			mod.cur=new.ou.lik.fn(cur.rates, cur.alpha, cur.regimes, cur.theta, cur.tre, ouch.dat)

    132. 

  132. 			mod.new=new.ou.lik.fn(new.rates, new.alpha, new.regimes, new.theta, new.tre, ouch.dat)

    133. 

  133. #			mod.new=try(new.ou.lik.fn(new.rates, new.alpha, new.regimes, new.theta, new.tre, ouch.dat),silent=TRUE)

    134. 

  134. 		} else if(cur.model=="BM"){

    135. 

  135. ## BM IMPLEMENTATION ##

    136. 

  136. 			if (runif(1) < (2 * probMergeSplit)) {

    137. 

  137. 				nr=split.or.merge(cur.delta.rates, cur.rates, ape.tre, theta=FALSE)

    138. 

  138. 				new.rates=nr$new.values

    139. 

  139. 				new.delta.rates=nr$new.delta

    140. 

  140. 				new.root=cur.root

    141. 

  141. 				nRcatProp=nRcatProp+1

    142. 

  142. 				lnHastingsRatio=nr$lnHastingsRatio

    143. 

  143. 				lnPriorRatio=nr$lnPriorRatio

    144. 

  144. 			} else { 

    145. 

  145. 				if(runif(1)<probRoot) { # adjust root

    146. 

  146. 					new.root=adjust.value(cur.root)

    147. 

  147. 					new.rates=cur.rates

    148. 

  148. 					new.delta.rates=cur.delta.rates

    149. 

  149. 					nRootProp=nRootProp+1

    150. 

  150. 				} else { # adjust rates

    151. 

  151. 					new.rates=adjust.rate(cur.rates)

    152. 

  152. 					new.delta.rates=cur.delta.rates

    153. 

  153. 					new.root=cur.root

    154. 

  154. 					nRateProp=nRateProp+1

    155. 

  155. 				}

    156. 

  156. 			}

    157. 

  157. 			

    158. 

  158. 			cur.tre=apply.BMM.to.apetree(ape.tre, cur.rates)

    159. 

  159. 			new.tre=apply.BMM.to.apetree(ape.tre, new.rates)

    160. 

  160. 			mod.cur=new.bm.lik.fn(cur.rates, cur.root, cur.tre, orig.dat)

    161. 

  161. 			mod.new=new.bm.lik.fn(new.rates, new.root, new.tre, orig.dat)

    162. 

  162. #			mod.new=try(new.bm.lik.fn(new.rates, new.root, new.tre, ouch.dat),silent=TRUE)

    163. 

  163. 		}

    164. 

  164. 	

    165. 

  165. 		if(!inherits(mod.new, "try-error")) {

    166. 

  166. 			lnLikelihoodRatio = mod.new$lnL - mod.cur$lnL

    167. 

  167. 		} else {

    168. 

  168. 			lnLikelihoodRatio = -Inf

    169. 

  169. 			mod.new$lnL=NaN

    170. 

  170. 		}

    171. 

  171. 		r=assess.lnR((heat * lnLikelihoodRatio + heat * lnPriorRatio + lnHastingsRatio)->lnR)

    172. 

  172. 

    173. 

  173. 		if (runif(1) <= r$r) {			# adopt proposal

    174. 

  174. 			if(cur.model=="OU") {

    175. 

  175. 				decision="adopt"

    176. 

  176. 				cur.alpha <- new.alpha

    177. 

  177. 				cur.root.theta <- new.root.theta

    178. 

  178. 				cur.regimes <- new.regimes

    179. 

  179. 				cur.theta <- new.theta

    180. 

  180. 				cur.delta.theta <- new.delta.theta

    181. 

  181. 			} else {

    182. 

  182. 				decision="adopt"

    183. 

  183. 				cur.root <- new.root

    184. 

  184. 			}

    185. 

  185. 			cur.rates <- new.rates

    186. 

  186. 			cur.delta.rates <- new.delta.rates

    187. 

  187. 			curr.lnL <- mod.new$lnL

    188. 

  188. 			cur.tre <- new.tre

    189. 

  189. 		} else {						# deny proposal	

    190. 

  190. 			decision="reject"

    191. 

  191. 			curr.lnL <- mod.cur$lnL 

    192. 

  192. 		}

    193. 

  193. 		

    194. 

  194. 		if(i==1) cat(paste("gen", "cur.lnL", "new.lnL", "lnR", sQuote(decision), sep="\t\t"),"\n")

    195. 

  195. 		cat(paste(i, sprintf("%.2f", mod.cur$lnL), sprintf("%.2f", mod.new$lnL), sprintf("%.2f", lnR), sQuote(decision), sep="\t\t"),"\n")

    196. 

  196. 

    197. 

  197. 		if(r$error) generate.error.message(i, mod.cur, mod.new, lnR, errorLog)

    198. 

  198. 		

    199. 

  199. 		l[i,]<-c(curr.lnL, sum(cur.delta.rates)+1, cur.alpha, length(table(cur.regimes)))

    200. 

  200. 		endparms=list(rates=cur.rates, cats=sum(cur.delta.rates)+1, root=cur.root, alph=cur.alpha, reg=length(table(cur.regimes)), theta=mean(cur.theta))

    201. 

  201. 		cOK=tally.mcmc.parms(startparms, endparms, cOK)

    202. 

  202. 		

    203. 

  203. 		bundled.parms=list(gen=i, mrate=exp(mean(log(cur.rates))), cats=sum(cur.delta.rates)+1, root=cur.root, alpha=cur.alpha, reg=length(table(cur.regimes)), theta=mean(cur.theta), lnL=curr.lnL)

    204. 

  204. 		if(i%%sampleFreq==0) generate.log(bundled.parms, cur.model, file=runLog)

    205. 

  205. 	}

    206. 

  206. 	

    207. 

  207. # End rjMCMC

    208. 

  208. 	close(errorLog)

    209. 

  209. 	close(runLog)

    210. 

  210. 	summarize.run(cOK, (cProp<-c(nRateProp, nRcatProp, nRootProp, nAlphaProp, nRegimesProp, nThetaProp)), cur.model)	

    211. 

  211. 	if(cur.model=="BM") {

    212. 

  212. 		return(list(r.cat=l[,2], lnL=l[,1]))

    213. 

  213. 	} else {

    214. 

  214. 		return(list(r.cat=l[,2], r.alpha=l[,3], r.regimes=l[,4], lnL=l[,1]))

    215. 

  215. 	}

    216. 

  216. }

    217. 

  217. 

    218. 

  218. 

    219. 

  219. ## AUXILIARY FUNCTIONS

    220. 

  220. 

    221. 

  221. 

    222. 

  222. new.ou.lik.fn <- function(rates, alpha, regimes, theta, ape.tre, ouch.dat) { # from OUCH 2.7-1:::hansen()

    223. 

  223. # mod 09.13.2010 JM Eastman

    224. 

  224. 	ouch.tre=ape2ouch(ape.tre, scale=FALSE)

    225. 

  225. 	ouch.dat$regimes=as.factor(regimes)

    226. 

  226. 	theta.tmp=as.numeric(levels(as.factor(theta)))

    227. 

  227. 	theta=theta.tmp[order(match(theta.tmp,theta))]

    228. 

  228. 	alpha.ouch=as.matrix(sqrt(alpha))

    229. 

  229. 	beta=ouch:::regime.spec(ouch.tre, ouch.dat['regimes'])

    230. 

  230. 	dat=do.call(c,lapply(ouch.dat['trait'],function(y)y[ouch.tre@term]))

    231. 

  231. 	names(dat)=ouch.dat$labels[Ntip(ape.tre):nrow(ouch.dat)]

    232. 

  232. 	n <- length(dat)

    233. 

  233. 	otree.df=as(ouch.tre,"data.frame")

    234. 

  234. 	ordering=otree.df[n:nrow(otree.df),'labels']

    235. 

  235. 	ev <- eigen(alpha.ouch,symmetric=TRUE)

    236. 

  236. 	w <- .Call(ouch:::ouch_weights,object=ouch.tre,lambda=ev$values,S=ev$vectors,beta=beta)

    237. 

  237. 	v <- geiger:::ouMatrix(vcv.phylo(ape.tre), alpha)

    238. 

  238. 	v.match <- match(row.names(v),ordering)

    239. 

  239. 	v <- v[v.match,v.match]

    240. 

  240. 		## GEIGER call does not assume ultrametricity (and thus allows different 'rates' under BM)

    241. 

  241. 		# -old-	v <- .Call(ouch:::ouch_covar,object=ouch.tre,lambda=ev$values,S=ev$vectors,sigma.sq=1)

    242. 

  242. 	

    243. 

  243. 	y <- matrix(theta,ncol=1)

    244. 

  244. 	e <- w%*%y-dat

    245. 

  245. 	dim(y) <- dim(y)[1]

    246. 

  246. 	dim(e) <- n

    247. 

  247. 	q <- e%*%solve(v,e)

    248. 

  248. 	det.v <- determinant(v,logarithm=TRUE)

    249. 

  249. 	if (det.v$sign!=1)stop("ou.lik.fn error: non-positive determinant",call.=FALSE)

    250. 

  250. 	dev <- n*log(2*pi)+as.numeric(det.v$modulus)+q[1,1]

    251. 

  251. 	list(

    252. 

  252. 		 theta=theta,

    253. 

  253. 		 weight=w,

    254. 

  254. 		 vcov=v,

    255. 

  255. 		 resids=e,

    256. 

  256. 		 lnL=-0.5*dev

    257. 

  257. 		 )

    258. 

  258. }

    259. 

  259. 

    260. 

  260. new.bm.lik.fn <- function(rates, root, ape.tre, orig.dat) { # from OUCH 2.7-1:::brown()	

    261. 

  261. # mod 09.16.2010 JM Eastman

    262. 

  262. 

    263. 

  263. 	tree=ape.tre

    264. 

  264. 	data=orig.dat

    265. 

  265. 

    266. 

  266. 	nterm=Ntip(tree)

    267. 

  267. 	nchar=length(data)

    268. 

  268. 	dat=data[match(tree$tip.label,names(data))]

    269. 

  269. 

    270. 

  270. 	nchar <- 1

    271. 

  271. 	w <- matrix(data=1,nrow=nterm,ncol=1)

    272. 

  272. 	n <- length(dat)

    273. 

  273. 

    274. 

  274. 	b <- vcv.phylo(ape.tre)

    275. 

  275. 

    276. 

  276. 	y <- matrix(root)

    277. 

  277. 	e <- w%*%y-dat

    278. 

  278. 	q <- t(e)%*%solve(b,e)

    279. 

  279. 	v <- q/nterm

    280. 

  280. 	dev <- nchar*nterm*(1+log(2*pi))+nchar*log(det(b))+nterm*log(det(v))

    281. 

  281. 	list(

    282. 

  282. 		 root=root,

    283. 

  283. 		 lnL = -0.5*dev

    284. 

  284. 		 )	

    285. 

  285. }

    286. 

  286. 

    287. 

  287. split.or.merge <- function(cur.delta, cur.values, phy, theta=FALSE) { 

    288. 

  288. # mod 09.14.2010 JM Eastman

    289. 

  289. 	bb=cur.delta

    290. 

  290. 	vv=cur.values

    291. 

  291. 	names(vv)<-names(bb)<-phy$edge[,2]

    292. 

  292. 	new.bb=choose.one(bb)

    293. 

  293. 	new.vv=vv

    294. 

  294. 	

    295. 

  295. 	s=names(new.bb[bb!=new.bb])

    296. 

  296. 	all.shifts=as.numeric(names(bb[bb>0]))

    297. 

  297. 	all.D=get.descendants.of.node(s, phy)

    298. 

  298. 	untouchable=unlist(lapply(all.shifts[all.shifts>s], function(x)get.descendants.of.node(x,phy)))

    299. 

  299. 	remain.unchanged=union(all.shifts, untouchable)

    300. 

  300. 	

    301. 

  301. 	marker=match(s, names(new.vv))

    302. 

  302. 	nn=length(vv)

    303. 

  303. 	K=length(table(vv))

    304. 

  304. 	N=Ntip(phy)

    305. 

  305. 	

    306. 

  306. 	if(sum(new.bb)>sum(bb)) {			# add transition: SPLIT

    307. 

  307. 		if(theta) {							# assign new theta to current node

    308. 

  308. 			new.vv[marker] <- nr <- adjust.value(new.vv[marker])

    309. 

  309. 		} else {							# assign new rate to current node

    310. 

  310. 			new.vv[marker] <- nr <- adjust.rate(new.vv[marker])

    311. 

  311. 		}

    312. 

  312. 		if(length(remain.unchanged)==0) {	# assign new value to all descendants

    313. 

  313. 			new.vv[match(all.D,names(new.vv))] = nr

    314. 

  314. 		} else {							# assign new value to all 'open' descendants 

    315. 

  315. 			new.vv[match(all.D[!(all.D%in%remain.unchanged)],names(new.vv))] = nr

    316. 

  316. 		}

    317. 

  317. 		lnHastingsRatio = log((K+1)/(2*N-2-K)) ### from Drummond and Suchard 2010: where N is tips, K is number of local parms in tree

    318. 

  318. 		lnPriorRatio = log(ppois(K+1,log(2),lower.tail=FALSE)/ppois(K,log(2),lower.tail=FALSE))

    319. 

  319. 	} else {							# drop transition: MERGE

    320. 

  320. 		anc = get.ancestor.of.node(s, phy)

    321. 

  321. 		if(!is.root(anc, phy)) {			# substitute ancestral value for current

    322. 

  322. 			new.vv[match(s, names(new.vv))] <- nr <- new.vv[match(anc, names(new.vv))] 

    323. 

  323. 		} else {							

    324. 

  324. 			if(theta) {						# assign new theta to first descendant of root

    325. 

  325. 				new.vv[marker] <- nr <- adjust.value(new.vv[marker])

    326. 

  326. 			} else {						# assign new rate to first descendant of root

    327. 

  327. 				new.vv[marker] <- nr <- adjust.rate(new.vv[marker])

    328. 

  328. 			}		

    329. 

  329. 		}

    330. 

  330. 		if(length(remain.unchanged)==0) {	# assign new value to all descendants

    331. 

  331. 			new.vv[match(all.D, names(new.vv))] = nr

    332. 

  332. 		} else {							# assign new value to all 'open' descendants

    333. 

  333. 			new.vv[match(all.D[!(all.D%in%remain.unchanged)],names(new.vv))] = nr

    334. 

  334. 		}

    335. 

  335. 		lnHastingsRatio = log((2*N-2-K+1)/K) ### from Drummond and Suchard 2010: where N is tips, K is number of local parms in tree

    336. 

  336. 		lnPriorRatio = log(ppois(K-1,log(2),lower.tail=FALSE)/ppois(K,log(2),lower.tail=FALSE))

    337. 

  337. 		

    338. 

  338. 	}

    339. 

  339. 	return(list(new.values=new.vv, new.delta=new.bb, lnHastingsRatio=lnHastingsRatio, lnPriorRatio=lnPriorRatio))

    340. 

  340. }

    341. 

  341. 

    342. 

  342. assign.root.theta <- function(cur.theta, phy) {

    343. 

  343. 	root=phy$edge[1,1]

    344. 

  344. 	desc=phy$edge[which(phy$edge[,1]==root),2]

    345. 

  345. 	root.theta=cur.theta[sample(which(phy$edge[,2]==desc),1)]

    346. 

  346. 	root.theta

    347. 

  347. }

    348. 

  348. 

    349. 

  349. get.descendants.of.node <- function(node, phy) {

    350. 

  350. 	storage=list()

    351. 

  351. 	if(node%in%phy$edge[,1]) {

    352. 

  352. 		desc=c(sapply(node, function(x) {return(phy$edge[which(phy$edge[,1]==x),2])}))

    353. 

  353. 		while(any(desc%in%phy$edge[,1])) {

    354. 

  354. 			desc.true=desc[which(desc%in%phy$edge[,1])]

    355. 

  355. 			desc.desc=lapply(desc.true, function(x) return(get.desc.of.node(x, phy)))

    356. 

  356. 			

    357. 

  357. 			storage=list(storage,union(desc,desc.desc))

    358. 

  358. 			

    359. 

  359. 			desc=unlist(desc.desc)

    360. 

  360. 		}

    361. 

  361. 	}

    362. 

  362. 	return(unique(unlist(storage)))

    363. 

  363. }

    364. 

  364. 

    365. 

  365. 

    366. 

  366. 

    367. 

  367. get.desc.of.node <- function(node, phy) {

    368. 

  368. 	return(phy$edge[which(phy$edge[,1]==node),2])

    369. 

  369. }

    370. 

  370. 

    371. 

  371. 		

    372. 

  372. get.ancestor.of.node<-function(node, phy) {

    373. 

  373. 	anc=phy$edge[which(phy$edge[,2]==node),1]

    374. 

  374. 	anc

    375. 

  375. }

    376. 

  376. 

    377. 

  377. choose.one <- function(cur.delta)

    378. 

  378. {

    379. 

  379. 	bb=cur.delta

    380. 

  380. 	s=sample(1:length(bb),1)

    381. 

  381. 	bb[s]=1-bb[s]

    382. 

  382. 	bb

    383. 

  383. }

    384. 

  384. 

    385. 

  385. is.root<-function(node,phy) {

    386. 

  386. 	if(node==phy$edge[1,1]) return(TRUE) else return(FALSE)

    387. 

  387. }

    388. 

  388. 

    389. 

  389. assess.lnR <- function(lnR) {

    390. 

  390. 	if(is.na(lnR) || abs(lnR)==Inf) {

    391. 

  391. 		error=TRUE

    392. 

  392. 		r=0

    393. 

  393. 	} else {

    394. 

  394. 		if(lnR < -100) {

    395. 

  395. 			r=0 

    396. 

  396. 		} else if(lnR > 0) {

    397. 

  397. 			r=1 

    398. 

  398. 		} else {

    399. 

  399. 			r=exp(lnR)

    400. 

  400. 		}

    401. 

  401. 		error=FALSE

    402. 

  402. 	}

    403. 

  403. 	return(list(r=r, error=error))

    404. 

  404. }

    405. 

  405. 

    406. 

  406. apply.BMM.to.apetree<-function(apetree, rates) {

    407. 

  407. 	apetree$edge.length=apetree$edge.length*rates

    408. 

  408. 	return(apetree)

    409. 

  409. }

    410. 

  410. 

    411. 

  411. adjust.value <- function(value) {

    412. 

  412. 	vv=value

    413. 

  413. 	rch <- runif(1, min = -2, max = 2)

    414. 

  414. 	vv <- vv + rch

    415. 

  415. 	vv

    416. 

  416. }

    417. 

  417. 

    418. 

  418. adjust.rate <- function(rate) {

    419. 

  419. 	rr=rate

    420. 

  420. 	r=log(rr)

    421. 

  421. 	rch=runif(1, min=-2, max=2)

    422. 

  422. 	nr=r+rch

    423. 

  423. 	rr=exp(nr)

    424. 

  424. 	rr

    425. 

  425. }

    426. 

  426. 	

    427. 

  427. prepare.ouchdata <- function(phy, data) {

    428. 

  428. 	td <- treedata(phy, data, sort = T)	

    429. 

  429. 	ouch.tre <- ape2ouch(td$phy->ape.tre, scale=FALSE)		

    430. 

  430. 	ouch.dat=data.frame(as.numeric(ouch.tre@nodes), as.character(ouch.tre@nodelabels))

    431. 

  431. 	names(ouch.dat)=c("nodes","labels")

    432. 

  432. 	ouch.dat[ouch.dat[,2]=="",2]=NA

    433. 

  433. 	ouch.dat$trait=NA

    434. 

  434. 	mM=match(ouch.dat$labels,names(data))

    435. 

  435. 	for(i in 1:nrow(ouch.dat)){

    436. 

  436. 		if(!is.na(mM[i])){

    437. 

  437. 			ouch.dat$trait[i]=data[mM[i]]	

    438. 

  438. 		}

    439. 

  439. 	}

    440. 

  440. 	return(list(ouch.tre=ouch.tre, ouch.dat=ouch.dat, ape.tre=td$phy, orig.dat=td$data[,1]))

    441. 

  441. }

    442. 

  442. 

    443. 

  443. summarize.run<-function(cOK, cProp, cur.model) {

    444. 

  444. 	names(cProp)<-names(cOK)<-c("rates", "rcats", "root", "alpha", "sRegimes", "theta")

    445. 

  445. 	names(cProp)=paste("prop.",names(cProp),sep="")

    446. 

  446. 	names(cOK)=paste("okay.",names(cOK),sep="")

    447. 

  447. 	cat("\n")

    448. 

  448. 	if(cur.model=="BM") {

    449. 

  449. 		print(cProp[1:3])

    450. 

  450. 		print(cOK[1:3])

    451. 

  451. 	} else {

    452. 

  452. 		print(cProp[c(1,2,4,5,6)])

    453. 

  453. 		print(cOK[c(1,2,4,5,6)])

    454. 

  454. 	}

    455. 

  455. }

    456. 

  456. 

    457. 

  457. generate.log<-function(bundled.parms, cur.model, file, init=FALSE) {

    458. 

  458. #	bundled.parms=list(gen=i, mrate=mean(cur.rates), cats=max(cur.delta.rates), root=cur.root, alpha=cur.alpha, reg=max(cur.regimes), theta=mean(cur.theta), lnL=curr.lnL)

    459. 

  459. 	res=bundled.parms

    460. 

  460. 

    461. 

  461. 	if(init) {

    462. 

  462. 		if(cur.model=="BM") msg=paste("gen", "model", "mean.rate", "rates", "root", "lnL", sep="\t") else msg=paste("gen", "model", "mean.rate", "rates", "alpha", "regimes", "mean.theta", "lnL", sep="\t")

    463. 

  463. 	} else {

    464. 

  464. 		if(cur.model=="BM") {

    465. 

  465. 			msg<-paste(res$gen, sQuote(cur.model), sprintf("%.3f", res$mrate), res$cats, sprintf("%.3f", res$root), sprintf("%.3f", res$lnL),sep="\t")

    466. 

  466. 		} else {

    467. 

  467. 			msg<-paste(res$gen, sQuote(cur.model), sprintf("%.3f", res$mrate), res$cats, sprintf("%.4f", res$alpha), res$reg, sprintf("%.3f", res$theta), sprintf("%.3f", res$lnL),sep="\t")

    468. 

  468. 		}

    469. 

  469. 	}

    470. 

  470. 	write(msg, file=file, append=TRUE)

    471. 

  471. }

    472. 

  472. 

    473. 

  473. generate.error.message<-function(i, mod.cur, mod.new, lnR, errorLog, init=FALSE) {

    474. 

  474. 	if(init) {

    475. 

  475. 		initmsg = write(paste("gen", "curr.lnL", "new.lnL", "lnR", sep="\t"), file=errorLog)

    476. 

  476. 	} else {

    477. 

  477. 		write(paste(i, sprintf("%.3f", mod.cur$lnL), sprintf("%.3f", mod.new$lnL), sprintf("%.3f", lnR), sep="\t"), file=errorLog)

    478. 

  478. 	}

    479. 

  479. }

    480. 

  480. 

    481. 

  481. tally.mcmc.parms<-function(startparms, endparms, cOK) {

    482. 

  482. 	index=array(dim=(length(startparms)-1))

    483. 

  483. 	for(i in 1:(length(startparms)-1)) {

    484. 

  484. 		index[i]=all(startparms[[i]]==endparms[[i]])

    485. 

  485. 	}

    486. 

  486. 	if(!all(index)) {

    487. 

  487. 		cOK[max(which(index==FALSE))]=cOK[max(which(index==FALSE))]+1

    488. 

  488. 	}

    489. 

  489. 	if(length(startparms$theta) == length(endparms$theta)) {

    490. 

  490. 		if(!all(sort(startparms$theta)==sort(endparms$theta))) {

    491. 

  491. 			cOK[6]=cOK[6]+1

    492. 

  492. 		}

    493. 

  493. 	}

    494. 

  494. 	cOK

    495. 

  495. }
    496.