A yet untested code

Efrain Gonzalez
1 parent 556b97bfaf
Showing 1 changed file with 227 additions and 0 deletions Show diff stats
RMarkovBlanket.r
@@ -0,0 +1,227 @@
+#Efrain Gonzalez
+#7/25/2017
+#Code for Markov Blanket
+
+
+#The required libraries
+library(pryr)
+library(MASS)
+library(dplyr)
+library(tidyr)
+library(readr)
+library(stringr)
+
+
+#Have the user choose an original Dot file that they want to use
+DotFile <- file.choose() 
+TheDotP1 <- read_delim(DotFile,delim = "\t",col_names = FALSE) %>%
+	dplyr::filter(!grepl("->|[{}]",X1)) %>%
+	dplyr::filter(!grepl("Banjo",X1)) %>%
+	dplyr::filter(!grepl("labeljust",X1))
+counterP1 <- 1
+sizeDotP1 <- dim(TheDotP1)[1]
+NewDotP1 <- matrix("0",ncol = 2, nrow = sizeDotP1)
+for(counterP1 in 1:sizeDotP1){
+	coldataP1 <- str_trim(TheDotP1[counterP1,1]) %>%
+		as.character(.,stringsAsFactors = FALSE)
+	if(grepl("Banjo|labeljust|>",coldataP1)==FALSE){
+		NumberP1 <- strsplit(coldataP1," ") %>%
+			.[[1]]%>%
+			.[1]
+		VarNameP1 <- strsplit(coldataP1," ") %>%
+			.[[1]] %>%
+			.[2] %>%
+			strsplit(.,"\"") %>%
+			.[[1]] %>%
+			.[grep("^\\w|^\\d",.)]
+		NewDotP1[counterP1,1] <- VarNameP1
+		NewDotP1[counterP1,2] <- NumberP1
+	}
+	if(grepl("->",coldataP1) == TRUE){
+		break
+	}
+}
+
+
+TheDotP2 <- read_delim(DotFile,delim = "\t",col_names = FALSE) %>%
+	dplyr::filter(grepl("->",X1))
+counterP2 <- 1
+sizeDotP2 <- dim(TheDotP2)[1]
+NewDotP2 <- matrix("0",ncol = 2, nrow = sizeDotP2)
+for(counterP2 in 1:sizeDotP2){
+	coldataP2 <- str_trim(TheDotP2[counterP2,1]) %>%
+		as.character(.,stringsAsFactors = FALSE)
+	ParentNumP2 <- strsplit(coldataP2,"->") %>%
+		.[[1]]%>%
+		.[1]
+	ChildNumP2 <- strsplit(coldataP2,"->") %>%
+		.[[1]] %>%
+		.[2] %>%
+		strsplit(.,";") %>%
+		.[[1]] %>%
+		.[1]
+	NewDotP2[counterP2,1] <- ParentNumP2
+	NewDotP2[counterP2,2] <- ChildNumP2
+}
+
+colnames(NewDotP2) <- c("Parents","Children")
+
+#Matching numbers to variable names
+NewDotP2_2 <- NewDotP2
+for(i in 1:sizeDotP1){
+	#Where is the variable located within NewDotP2 (column one only)?
+	chngreq	<- grep(paste0("^",NewDotP1[i,2],"$"),NewDotP2_2[,1])
+	if(is.na(sum(chngreq)) == FALSE){
+		if(sum(chngreq) > 0){
+			NewDotP2_2[chngreq,1] <- gsub(paste0("^",NewDotP1[i,2],"$"),NewDotP1[i,1],NewDotP2_2[chngreq,1])
+		}
+	}
+	i <- i + 1
+}
+NewDotP2_2
+for(j in 1:sizeDotP1){
+	#Where is the variable located within NewDotP2 (column two only)?
+	chngreq	<- grep(paste0("^",NewDotP1[j,2],"$"),NewDotP2_2[,2])
+	if(is.na(sum(chngreq)) == FALSE){
+		if(sum(chngreq) > 0){
+			NewDotP2_2[chngreq,2] <- gsub(paste0("^",NewDotP1[j,2],"$"),NewDotP1[j,1],NewDotP2_2[chngreq,2])
+		}
+	}
+	j <- j + 1
+}
+
+lrgMarkov <- dim(NewDotP2_2)[1]
+MBlanky <- function(MarkovDegree = lrgMarkov, VariableStartName = "Alzheimer", VariableEndName = "Age"){
+	#Finding the Parents and Children
+	d <- 1
+	AllNamList1 <- vector("list",length = 3)
+	#AllNamList <- vector("list", length = 3)
+		##Finding the Parents of the Parents
+	##A list of lists
+	##outer set by the degree of the Markov blanket
+	AllVarList <- vector("list",length = 3)
+	for(d in 1:MarkovDegree){
+		colnames(NewDotP2_2) <- NULL
+		##Which variable are you looking for?
+		##This is the VariableEndName
+		if(d == 1){
+			##Finding the Parents for the variable
+			LocPofVar <- grep(VariableStartName,NewDotP2_2[,2])
+			PofVar <- NewDotP2_2[LocPofVar,1]
+			AllNamList1[[1]] <- PofVar
+			
+			##Finding the Children for the variable
+			LocCofVar <- grep(VariableStartName,NewDotP2_2[,1])
+			CofVar <- NewDotP2_2[LocCofVar,2]
+			AllNamList1[[2]] <- CofVar
+			
+			##Finding the Co-Parents of the Children for the variable
+			NumofChild <- length(CofVar)
+			if(NumofChild > 0){
+				##Creating a list of the Co-Parents for each of the children
+				##list size is based on the amount of Children
+				COPlist <- vector("character", length = 0)
+				nc <- 1
+				for(nc in 1:NumofChild){
+					LocCOPofVar <- grep(CofVar[nc],NewDotP2_2[,2])
+					COPofVar <- NewDotP2_2[LocCOPofVar,1]
+					if(grepl(COPofVar,VariableStartName)){
+						next
+					} else{
+						#COPlist[[nc]] <- COPofVar[COPofVar!=VariableStartName]
+						COPlist <- append(COPlist,COPofVar)
+					}
+					nc <- nc + 1
+				} 
+			} else {
+				##Making COPlist empty
+				COPlist <- vector("character",length = 0)
+			}
+			AllNamList1[[3]] <- COPlist
+			AllVarList[[1]] <- AllNamList1
+		} else if(d > 1){
+			##inner set by the length of the previous AllVarlist we are working on
+			lPreVList <- length(AllVarList[[d-1]])
+			ef <- 1
+			##PCCP will eventually equal the total size that we expect for the iteration (#d)
+			## which is just lPreVList * 3
+			PCCP <- 1
+			newsize <- (lPreVList * 3)
+			if(d > 3){
+				AllVarList[[d]] <- vector("list")
+			}
+			for(ef in 1:lPreVList){
+				##Finding the Parents
+				NumofVars <- length(AllVarList[[d-1]][[ef]])
+				if(NumofVars > 0){
+					##Creating a list of the Parents for each of the Variables
+					##list size is based on the amount of Previous Variables
+					PofVlist <- vector("character", length = 0)
+					np <- 1
+					for(np in 1:NumofVars){
+						LocPofVar <- grep(AllVarList[[d-1]][[ef]][np],NewDotP2_2[,2])
+						PofVar <- NewDotP2_2[LocPofVar,1]
+						PofVlist <- append(PofVlist,PofVar)
+						np <- np + 1
+					}
+				} else {
+					##Making COPlist empty
+					PofVlist <- vector("character",length = 0)
+				}
+				AllVarList[[d]][[PCCP]] <- PofVlist 
+				PCCP <- PCCP + 1
+				
+				##Finding the Children
+				if(NumofVars > 0){
+					##Creating a list of the Children for each of the Previous Parents
+					##list size is based on the amount of Previous Parents
+					CofVlist <- vector("character", length = 0)
+					np <- 1
+					for(np in 1:NumofVars){
+						LocCofVar <- grep(AllVarList[[d-1]][[ef]][np],NewDotP2_2[,1])
+						CofVar <- NewDotP2_2[LocCofVar,2]
+						#if(grepl(VariableStartName,CofVar)){
+						#	next
+						#} else{
+							CofVlist <- append(CofVlist,CofVar)
+						#}
+						np <- np + 1
+					}
+				} else {
+					##Making CofPlist empty
+					CofVlist <- vector("character",length = 0)
+				}
+				AllVarList[[d]][[PCCP]] <- CofVlist
+				PCCP <- PCCP + 1
+				
+				##Finding the Co-Parents
+				NumofCVars <- length(CofVlist)
+				if(NumofCVars > 0){
+					ncp <- 1
+					CPofClist <- vector("character",length = 0)
+					for(ncp in 1:NumofCVars){
+						LocCPofCVar <- grep(CofVlist[ncp],NewDotP2_2[,2])
+						CPofCVar <- NewDotP2_2[LocCPofCVar,1]
+						CPofClist <- append(CPofClist,CPofCVar)
+						ncp <- ncp + 1
+					}
+				
+				} else {
+					##Making COPlist empty
+					CPofClist <- vector("character",length = 0)
+				}
+				AllVarList[[d]][[PCCP]] <- CPofClist
+				PCCP <- PCCP + 1		
+				ef <- ef + 1
+			}
+		}
+		##Stop if you have found the VariableEndName value
+		if(sum(grepl(VariableEndName,AllVarList)) > 0){
+			break
+		}
+		d <- d + 1
+	}
+	##The Markov Degree is that found below
+	d
+}
+#Now use the command MBlanky() with the appropriate settings
 \ No newline at end of file
...	...	@@ -0,0 +1,227 @@
	1	+#Efrain Gonzalez
	2	+#7/25/2017
	3	+#Code for Markov Blanket
	4	+
	5	+
	6	+#The required libraries
	7	+library(pryr)
	8	+library(MASS)
	9	+library(dplyr)
	10	+library(tidyr)
	11	+library(readr)
	12	+library(stringr)
	13	+
	14	+
	15	+#Have the user choose an original Dot file that they want to use
	16	+DotFile <- file.choose()
	17	+TheDotP1 <- read_delim(DotFile,delim = "\t",col_names = FALSE) %>%
	18	+ dplyr::filter(!grepl("->\|[{}]",X1)) %>%
	19	+ dplyr::filter(!grepl("Banjo",X1)) %>%
	20	+ dplyr::filter(!grepl("labeljust",X1))
	21	+counterP1 <- 1
	22	+sizeDotP1 <- dim(TheDotP1)[1]
	23	+NewDotP1 <- matrix("0",ncol = 2, nrow = sizeDotP1)
	24	+for(counterP1 in 1:sizeDotP1){
	25	+ coldataP1 <- str_trim(TheDotP1[counterP1,1]) %>%
	26	+ as.character(.,stringsAsFactors = FALSE)
	27	+ if(grepl("Banjo\|labeljust\|>",coldataP1)==FALSE){
	28	+ NumberP1 <- strsplit(coldataP1," ") %>%
	29	+ .[[1]]%>%
	30	+ .[1]
	31	+ VarNameP1 <- strsplit(coldataP1," ") %>%
	32	+ .[[1]] %>%
	33	+ .[2] %>%
	34	+ strsplit(.,"\"") %>%
	35	+ .[[1]] %>%
	36	+ .[grep("^\\w\|^\\d",.)]
	37	+ NewDotP1[counterP1,1] <- VarNameP1
	38	+ NewDotP1[counterP1,2] <- NumberP1
	39	+ }
	40	+ if(grepl("->",coldataP1) == TRUE){
	41	+ break
	42	+ }
	43	+}
	44	+
	45	+
	46	+TheDotP2 <- read_delim(DotFile,delim = "\t",col_names = FALSE) %>%
	47	+ dplyr::filter(grepl("->",X1))
	48	+counterP2 <- 1
	49	+sizeDotP2 <- dim(TheDotP2)[1]
	50	+NewDotP2 <- matrix("0",ncol = 2, nrow = sizeDotP2)
	51	+for(counterP2 in 1:sizeDotP2){
	52	+ coldataP2 <- str_trim(TheDotP2[counterP2,1]) %>%
	53	+ as.character(.,stringsAsFactors = FALSE)
	54	+ ParentNumP2 <- strsplit(coldataP2,"->") %>%
	55	+ .[[1]]%>%
	56	+ .[1]
	57	+ ChildNumP2 <- strsplit(coldataP2,"->") %>%
	58	+ .[[1]] %>%
	59	+ .[2] %>%
	60	+ strsplit(.,";") %>%
	61	+ .[[1]] %>%
	62	+ .[1]
	63	+ NewDotP2[counterP2,1] <- ParentNumP2
	64	+ NewDotP2[counterP2,2] <- ChildNumP2
	65	+}
	66	+
	67	+colnames(NewDotP2) <- c("Parents","Children")
	68	+
	69	+#Matching numbers to variable names
	70	+NewDotP2_2 <- NewDotP2
	71	+for(i in 1:sizeDotP1){
	72	+ #Where is the variable located within NewDotP2 (column one only)?
	73	+ chngreq <- grep(paste0("^",NewDotP1[i,2],"$"),NewDotP2_2[,1])
	74	+ if(is.na(sum(chngreq)) == FALSE){
	75	+ if(sum(chngreq) > 0){
	76	+ NewDotP2_2[chngreq,1] <- gsub(paste0("^",NewDotP1[i,2],"$"),NewDotP1[i,1],NewDotP2_2[chngreq,1])
	77	+ }
	78	+ }
	79	+ i <- i + 1
	80	+}
	81	+NewDotP2_2
	82	+for(j in 1:sizeDotP1){
	83	+ #Where is the variable located within NewDotP2 (column two only)?
	84	+ chngreq <- grep(paste0("^",NewDotP1[j,2],"$"),NewDotP2_2[,2])
	85	+ if(is.na(sum(chngreq)) == FALSE){
	86	+ if(sum(chngreq) > 0){
	87	+ NewDotP2_2[chngreq,2] <- gsub(paste0("^",NewDotP1[j,2],"$"),NewDotP1[j,1],NewDotP2_2[chngreq,2])
	88	+ }
	89	+ }
	90	+ j <- j + 1
	91	+}
	92	+
	93	+lrgMarkov <- dim(NewDotP2_2)[1]
	94	+MBlanky <- function(MarkovDegree = lrgMarkov, VariableStartName = "Alzheimer", VariableEndName = "Age"){
	95	+ #Finding the Parents and Children
	96	+ d <- 1
	97	+ AllNamList1 <- vector("list",length = 3)
	98	+ #AllNamList <- vector("list", length = 3)
	99	+ ##Finding the Parents of the Parents
	100	+ ##A list of lists
	101	+ ##outer set by the degree of the Markov blanket
	102	+ AllVarList <- vector("list",length = 3)
	103	+ for(d in 1:MarkovDegree){
	104	+ colnames(NewDotP2_2) <- NULL
	105	+ ##Which variable are you looking for?
	106	+ ##This is the VariableEndName
	107	+ if(d == 1){
	108	+ ##Finding the Parents for the variable
	109	+ LocPofVar <- grep(VariableStartName,NewDotP2_2[,2])
	110	+ PofVar <- NewDotP2_2[LocPofVar,1]
	111	+ AllNamList1[[1]] <- PofVar
	112	+
	113	+ ##Finding the Children for the variable
	114	+ LocCofVar <- grep(VariableStartName,NewDotP2_2[,1])
	115	+ CofVar <- NewDotP2_2[LocCofVar,2]
	116	+ AllNamList1[[2]] <- CofVar
	117	+
	118	+ ##Finding the Co-Parents of the Children for the variable
	119	+ NumofChild <- length(CofVar)
	120	+ if(NumofChild > 0){
	121	+ ##Creating a list of the Co-Parents for each of the children
	122	+ ##list size is based on the amount of Children
	123	+ COPlist <- vector("character", length = 0)
	124	+ nc <- 1
	125	+ for(nc in 1:NumofChild){
	126	+ LocCOPofVar <- grep(CofVar[nc],NewDotP2_2[,2])
	127	+ COPofVar <- NewDotP2_2[LocCOPofVar,1]
	128	+ if(grepl(COPofVar,VariableStartName)){
	129	+ next
	130	+ } else{
	131	+ #COPlist[[nc]] <- COPofVar[COPofVar!=VariableStartName]
	132	+ COPlist <- append(COPlist,COPofVar)
	133	+ }
	134	+ nc <- nc + 1
	135	+ }
	136	+ } else {
	137	+ ##Making COPlist empty
	138	+ COPlist <- vector("character",length = 0)
	139	+ }
	140	+ AllNamList1[[3]] <- COPlist
	141	+ AllVarList[[1]] <- AllNamList1
	142	+ } else if(d > 1){
	143	+ ##inner set by the length of the previous AllVarlist we are working on
	144	+ lPreVList <- length(AllVarList[[d-1]])
	145	+ ef <- 1
	146	+ ##PCCP will eventually equal the total size that we expect for the iteration (#d)
	147	+ ## which is just lPreVList * 3
	148	+ PCCP <- 1
	149	+ newsize <- (lPreVList * 3)
	150	+ if(d > 3){
	151	+ AllVarList[[d]] <- vector("list")
	152	+ }
	153	+ for(ef in 1:lPreVList){
	154	+ ##Finding the Parents
	155	+ NumofVars <- length(AllVarList[[d-1]][[ef]])
	156	+ if(NumofVars > 0){
	157	+ ##Creating a list of the Parents for each of the Variables
	158	+ ##list size is based on the amount of Previous Variables
	159	+ PofVlist <- vector("character", length = 0)
	160	+ np <- 1
	161	+ for(np in 1:NumofVars){
	162	+ LocPofVar <- grep(AllVarList[[d-1]][[ef]][np],NewDotP2_2[,2])
	163	+ PofVar <- NewDotP2_2[LocPofVar,1]
	164	+ PofVlist <- append(PofVlist,PofVar)
	165	+ np <- np + 1
	166	+ }
	167	+ } else {
	168	+ ##Making COPlist empty
	169	+ PofVlist <- vector("character",length = 0)
	170	+ }
	171	+ AllVarList[[d]][[PCCP]] <- PofVlist
	172	+ PCCP <- PCCP + 1
	173	+
	174	+ ##Finding the Children
	175	+ if(NumofVars > 0){
	176	+ ##Creating a list of the Children for each of the Previous Parents
	177	+ ##list size is based on the amount of Previous Parents
	178	+ CofVlist <- vector("character", length = 0)
	179	+ np <- 1
	180	+ for(np in 1:NumofVars){
	181	+ LocCofVar <- grep(AllVarList[[d-1]][[ef]][np],NewDotP2_2[,1])
	182	+ CofVar <- NewDotP2_2[LocCofVar,2]
	183	+ #if(grepl(VariableStartName,CofVar)){
	184	+ # next
	185	+ #} else{
	186	+ CofVlist <- append(CofVlist,CofVar)
	187	+ #}
	188	+ np <- np + 1
	189	+ }
	190	+ } else {
	191	+ ##Making CofPlist empty
	192	+ CofVlist <- vector("character",length = 0)
	193	+ }
	194	+ AllVarList[[d]][[PCCP]] <- CofVlist
	195	+ PCCP <- PCCP + 1
	196	+
	197	+ ##Finding the Co-Parents
	198	+ NumofCVars <- length(CofVlist)
	199	+ if(NumofCVars > 0){
	200	+ ncp <- 1
	201	+ CPofClist <- vector("character",length = 0)
	202	+ for(ncp in 1:NumofCVars){
	203	+ LocCPofCVar <- grep(CofVlist[ncp],NewDotP2_2[,2])
	204	+ CPofCVar <- NewDotP2_2[LocCPofCVar,1]
	205	+ CPofClist <- append(CPofClist,CPofCVar)
	206	+ ncp <- ncp + 1
	207	+ }
	208	+
	209	+ } else {
	210	+ ##Making COPlist empty
	211	+ CPofClist <- vector("character",length = 0)
	212	+ }
	213	+ AllVarList[[d]][[PCCP]] <- CPofClist
	214	+ PCCP <- PCCP + 1
	215	+ ef <- ef + 1
	216	+ }
	217	+ }
	218	+ ##Stop if you have found the VariableEndName value
	219	+ if(sum(grepl(VariableEndName,AllVarList)) > 0){
	220	+ break
	221	+ }
	222	+ d <- d + 1
	223	+ }
	224	+ ##The Markov Degree is that found below
	225	+ d
	226	+}
	227	+#Now use the command MBlanky() with the appropriate settings
0	228	\ No newline at end of file