R/printInitialDescription.r

Defines functions printInitialDescription

#/******************************************************************************
#/******************************************************************************
# * SIENA: Simulation Investigation for Empirical Network Analysis
# *
# * Web: https://www.stats.ox.ac.uk/~snijders/siena
# *
# * File: printInitialDescription.r
# *
# * Description: This module contains the code to print details of the project.
# *
# *****************************************************************************/
##@printInitialDescription Reporting
printInitialDescription <- function(data, effects, modelName="Siena",
	getDocumentation=FALSE)
{
	##@initialNetworks internal printInitialDescription
	initialNetworks <- function()
	{
		Heading(2, outf,'Change in networks:')
		difficult <- FALSE
		for (net in which(types %in% c("oneMode", "bipartite")))
		{
			bipartite <- types[net] == "bipartite"
			if ((nOneMode + nBipartite) > 1)
			{
				Heading(3, outf, c("Network: ", netnames[net]))
			}

			Report(c("For the following statistics, missing values (if any)",
					"are not counted.\n"), outf)
			if (any(attr(data, "structural")))
			{
				Report(c("Further, structurally determined entries are",
						"treated as observed entries,\n"), outf)
				Report(c("except for calculating the distances, where missings",
						"and structurally determined entries\n"), outf)
				Report(c("at the beginning or end of a period",
						"are disregarded.\n"), outf)
				Report(c("This may imply that the columns 0 => 1 and 1 => 0",
						"do not add up to the distance column.\n"), outf)
			}
			## is this network symmetric?
			if (gpatts$symmetric[net])
			{
				Report(c(ifelse(nobs == 2, "Both", c("All ", nobs)),
						" observed networks are symmetric.\n"), outf)
				Report(c("Therefore, it is assumed that this is an",
						"analysis of an non-directed relation.\n"), outf)
			}
			Report("\nNetwork density indicators:\n", outf)
			density <- rep(NA, nobs)
			degree <- rep(NA, nobs)
			nties <- rep(NA, nobs)
			missings <- rep(NA, nobs)
			periodFromStart <- 0
			for (group in 1:nData)
			{
				j <- match(netnames[net], names(data[[group]]$depvars))
				if (is.na(j))
					stop("network names not consistent")
				depvar <- data[[group]]$depvars[[j]]
				atts <- attributes(depvar)
				subs <- 1:data[[group]]$observations + periodFromStart
				density[subs] <- atts$density
				if (any(atts$ones >= atts$nval, na.rm = TRUE))
				{
					difficult <- TRUE
				}
				degree[subs] <- atts$degree
				missings[subs] <-atts$missings
				nties[subs] <- atts$ones
				if (gpatts$symmetric[net])
				{
					nties <- nties / 2
				}
				periodFromStart <- periodFromStart + data[[group]]$observations
			}
			## now do the format
			tmp <- rbind(format(round(density, 3), nsmall=3, width=7),
				format(round(degree, 3), nsmall=3, width=7),
				format(nties, width=7),
				format(round(missings, 3), nsmall=3, width=7))
			startCol <- 1
			text <- c("observation time", "density", "average degree",
				"number of ties", "missing fraction")
			text <- c(format(text[1], width=24), format(text[-1], width=25))
			eols <- rep("\n", length(text))
			repeat
			{
				endCol <- startCol + 6
				endCol <- min(endCol, nobs)
				tmp3 <- rbind(format(startCol:endCol, width=7),
					tmp[, startCol:endCol, drop=FALSE])
				tmp2 <- cbind(text, tmp3, eols)
				Report(t(tmp2), sep="", outf)
				startCol <- startCol + 7
				if (startCol > nobs)
					break
			}
			averageOutDegree <- rep(NA, nData)
			for (group in 1:nData)
			{
				j <- match(netnames[net], names(data[[group]]$depvars))
				if (is.na(j))
					stop("network names not consistent")
				depvar <- data[[group]]$depvars[[j]]
				atts <- attributes(depvar)
				averageOutDegree[group] <- atts$"averageOutDegree"
			}
			Report("\n", outf)
			if (nData > 1)
			{
				Report("The average degrees over all waves are: \n ", outf)
				Report(paste(names(data), round(averageOutDegree, 3), "\n",
						sep=': '), outf)
				Report("\n", outf)
			}
			else
			{
				Report(c("The average degree is",
						round(averageOutDegree, 3), "\n"), outf)
			}
			Report("\n\n", outf)
			Report(c(ifelse(gpatts$symmetric[net], "Edge", "Tie"),
					"changes between subsequent observations:\n"), outf)
			valmin <- gpatts$netRanges[1, net]
			valmax <- gpatts$netRanges[2, net]
			tmp <- expand.grid(valmin:valmax, valmin:valmax)
			heads <- paste(tmp[, 2], "=> ", tmp[,1])
			Report(c(format(" periods", width=16),
					format(heads, width=10),
					ifelse(valmin == 0 && valmax == 1,
						"Distance Jaccard   Missing\n",
						"Distance     Missing\n")), sep = "", outf)
			periodFromStart <- 0

			for (group in 1:nData)
			{
				j <- match(netnames[net], names(data[[group]]$depvars))
				if (is.na(j))
					stop("network names not consistent")
				depvar <- data[[group]]$depvars[[j]]
				if (bipartite)
				{
					tmp <- getBipartiteStartingVals(depvar)
				}
				else
				{
					tmp <- getNetworkStartingVals(depvar)
				}
				atts <- attributes(depvar)
				matchange <- tmp$tmp[-(1:2), , drop=FALSE]
				## if symmetric: divide by 2
				if (gpatts$symmetric[net])
				{
					tmp$tmp <- tmp$tmp %/% 2
					matchange <- matchange %/% 2
				}
				for (per in 1:(atts$netdims[3] - 1))
				{
					ntot <- tmp$tmp["matcnt", per]
					if (gpatts$symmetric[net])
					{
						misd <- atts$netdims[1] * (atts$netdims[1] - 1) / 2 -
							ntot
					}
					else if (bipartite)
					{
						misd <- atts$netdims[1] * atts$netdims[2] - ntot
					}
					else
					{
						misd <- atts$netdims[1] * (atts$netdims[1] - 1) - ntot
					}
					if (valmin == 0 && valmax == 1)
					{
						jaccard <- format(round(matchange[4, per] /
								(matchange[4, per] +
									matchange[3, per] +
									matchange[2, per]), 3), nsmall=3,
							width=10)

						Report(c(format(per + periodFromStart, width=3),
								" ==> ",
								format(per + 1 + periodFromStart, width=3),
								format(matchange[, per], width=10),
								format(attr(depvar, "distance")[per],
									width=10),
								jaccard, format(misd, width=10), " (",
								round(100 * misd/(ntot + misd)), "%)\n"),
							sep="", outf)
					}
					else
					{
						Report(c(per + periodFromStart, " ==> ",
								format(per + 1 + periodFromStart, width=3),
								format(matchange[, per], width=10),
								format(attr(depvar,"distance")[per], width=10),
								format(misd, width=10), " (",
								round(100 * misd/(ntot + misd)), "%)\n"),
							sep="", outf)

					}
					if (valmin == 0 && valmax == 1)
					{
						if (matchange[4, per]*
							(matchange[2, per] + matchange[1, per])
						<
						matchange[2, per] *
							(matchange[3, per] + matchange[4, per]) )
						{
							Report(c("\nThis means that in period ", per,
									", proportionately fewer 1-ties stayed 1,\n",
									" than 0-ties became 1. A great reversal",
									" of the network pattern!\n",
									"For some model specifications this may",
									" lead to problems in estimation.\n"), outf)
						}
					}
				}
				periodFromStart <- periodFromStart + atts$netdims[3]
				Report("\n", outf)
			}
			if (gpatts$symmetric[net])
			{
				Report(c("The distances reported in the output file",
						"for conditional estimation\n",
						"for the network variable refer to the total",
						"symmetric adjacency matrix,\n",
						"and therefore are double the distance",
						"reported above.)\n\n"), outf)
			}
			if (!bipartite)
			{
				Report("Directed dyad Counts:\n", outf)
				if (valmin == 0 & valmax == 1)
				{
					Report(" observation    total    mutual    asymm.     null\n",
						outf)
				}
				else
				{
					Report(" observation    total \n", outf)
					heads <- expand.grid(valmin:valmax, valmin:valmax)
					heads <- heads[, c(2, 1)] ## reverse the column order
					heads <- heads[heads[,2] >= heads[,1], ] ## remove col2 < col1
					heads <- paste("(", heads[, 1], ",", heads[, 2], ")", sep="")
					Report(heads, outf)
					Report("\n", outf)
					## this one needs more work!
				}
				periodFromStart <- 0
				for (group in 1:nData)
				{
					j <- match(netnames[net], names(data[[group]]$depvars))
					if (is.na(j))
						stop("network names not consistent")
					depvar <- data[[group]]$depvars[[j]]
					atts <- attributes(depvar)

					for (per in 1:(atts$netdims[3]))
					{
						## find the dyad tables
						if (atts$sparse)
						{
							##  require(Matrix)
							mymat <- depvar[[per]]
							diag(mymat) <- 0
							mymat1 <- mymat@i
							mymat2 <- mymat@j
							mymat3 <- mymat@x
							if (any(duplicated(paste(mymat1, mymat2))))
								stop("sparse matrix has duplicate triples")
							mymat3[mymat3==10] <- 0
							mymat3[mymat3==11] <- 1
							mymat1 <- mymat1[is.na(mymat3) | mymat3 != 0]
							mymat2 <- mymat2[is.na(mymat3) | mymat3 != 0]
							mymat3 <- mymat3[is.na(mymat3) | mymat3 != 0]
							## for not 0/1 need to tabulate the equivs to mutuals
							## and asymms by
							## the non zero one.
							if (valmin == 0 && valmax ==1)
							{
								ij <- paste(mymat1[!is.na(mymat3)],
									mymat2[!is.na(mymat3)])
								ji <- paste(mymat2[!is.na(mymat3)],
									mymat1[!is.na(mymat3)])
								missij <- paste(mymat1[is.na(mymat3)],
									mymat2[is.na(mymat3)])
								missji <- paste(mymat2[is.na(mymat3)],
									mymat1[is.na(mymat3)])
								mutual <- sum(ij %in% ji) / 2
								## nondyads are ones where we have a link and
								## its partner is missing
								nondyads <- sum(ji %in% missij)
								asymm <- length(ij) - nondyads - mutual * 2
								missdyads <- sum(!(missij %in% missji)) +
									sum(missij %in% missji) / 2
								nulls <- atts$netdims[1] *
									(atts$netdims[2] - 1) / 2 -
									missdyads - mutual - asymm
								totDyad <- nulls + mutual + asymm
							}
						}
						else
						{
							mymat <- depvar[, , per]
							mymat[mymat == 10] <- 0
							mymat[mymat == 11] <- 1
							diag(mymat) <- NA
							tmymat <- t(mymat)
							# dyadTable <- table(mymat, t(mymat))
							# Changed to protect against zero rows or columns
							nulls <- sum((1 - mymat)*(1 - tmymat), na.rm=TRUE)
							asymm <- sum(mymat*(1 - tmymat), na.rm=TRUE) +
								sum((1 - mymat)*tmymat, na.rm=TRUE)
							mutual <- sum(mymat*tmymat, na.rm=TRUE)
							# diag(dyadTable) <- diag(dyadTable) / 2
							#if (valmin == 0 && valmax ==1)
							#{
							#    mutual <- dyadTable[2, 2]
							#    asymm <- dyadTable[2, 1]
							#    nulls <- dyadTable[1, 1]
							totDyad <- nulls + asymm + mutual
							#}
						}
						if (valmin == 0 && valmax == 1)
						{
							Report(c(format(per + periodFromStart, width=6),
									".", format(totDyad, width=14),
									format(mutual, width=9),
									format(asymm, width=10),
									format(nulls, width=10), "\n"),
								sep="", outf)
						}
					}
					periodFromStart <- periodFromStart + atts$netdims[3]
				}
			}
			if (difficult)
			{
				Report(c("There is a density equal to 0.0 or 1.0.\n",
						"This may lead to difficulties.\n"), outf)
			}
			Report("\n", outf)
			Report(c("Standard values for initial parameter values\n",
					"-------------------------------------------------\n\n"),
				sep = "", outf)
			myeff <- effects[effects$name == netnames[net],]
			myrate <- myeff[myeff$shortName == "Rate",]
			for (i in 1:nrow(myrate))
			{
				Report(c(format(myrate$effectName[i], width=35),
						format(round(myrate$initialValue[i], 4), nsmall=4,
							width=10), "\n"), outf)
			}
			myobj <- myeff[myeff$shortName == "density" &
				myeff$type == "eval",]
			if (nrow(myobj) > 0)
			{
				untrimmed <- myobj$untrimmedValue
				Report(c(format(myobj$effectName, width=46),
						format(round(untrimmed, 4), nsmall=4, width=10),
						"\n"), outf)
				if (untrimmed > 3)
				{
					Report(c("The initial parameter value is very low.",
							"It is truncated to -3. \n"), outf)
				}
				else
					if (untrimmed < -3)
					{
						Report(c("The initial parameter value is very high.",
								"It is truncated to 3.\n"), outf)
					}
			}
		}
		Report("\n", outf)
	}

	##@initialBehaviors internal printInitialDescription
	initialBehaviors <- function()
	{
		Report("\n", outf)
     		Heading(2, outf,"Dependent discrete actor variables:")
		if (nBehav == 1)
		{
			Report(c(netnames[types == "behavior"], "\n\n"), outf)
			Heading(3, outf, "Marginal distribution")
		}
		else
		{
			Heading(3, outf, "Marginal distributions")
		}
		for (net in which(types=="behavior"))
		{
			if (nBehav > 1)
			{
				Report(c("Dependent actor variable ", net, ": ",
						netnames[net], "\n"), sep="", outf)
			}
			Report(c(rep(" ", 3*nobs + 7), "Observations\n"), sep="", outf)
			Report(c(format("values", width=16),
					format(1 : nobs, width=6), "\n"), sep="", outf)
			Report(c(rep(" ", 16), rep("------", nobs), "\n"), sep="", outf)
			periodFromStart <- 0
			bRange <- gpatts$bRange[net]
			minval <- gpatts$behRange[, net][1]
			vals <- matrix(0, ncol=nobs, nrow=bRange + 1)
			missings <- rep(0, nobs)
			missingN <- rep(NA, nobs)
			for (group in 1:nData)
			{
				j <- match(netnames[net], names(data[[group]]$depvars))
				if (is.na(j))
					stop("network names not consistent")
				depvar <- data[[group]]$depvars[[j]]
				atts <- attributes(depvar)
				for (i in 1: atts$netdims[3])
				{
					mytab <- table(round(depvar[, 1, i]))
					vals[as.numeric(names(mytab)) + 1 - minval,
						periodFromStart + i] <- mytab
					# vals[factor(names(mytab)), periodFromStart + i] <- mytab
					missings[periodFromStart + i] <- sum(is.na(depvar[, 1, i]))
					missingN[periodFromStart + i] <- atts$netdims[1]
				}
				periodFromStart <- periodFromStart + atts$netdims[3]
			}
			for (i in 1:nrow(vals))
			{
				Report(c(format(i - 1 + minval, width=3), rep(" ", 13),
						format(vals[i, ], width=6), "\n"), sep="", outf)
			}
			if (gpatts$anyMissing[net])
			{
				Report(c(format("missing", width=16),
						format(missings, width=6), "\n\n"), sep="", outf)
				Report(c(format("(fraction missing"),
						format(round(missings / missingN, 2), nsmall=2,
							width=6), "  )\n"), sep="", outf)
			}
			else
			{
				Report("No missings\n", outf)
			}

		}
		Report("\n\n", outf)
		Heading(3, outf, "Changes")
		for (net in which(types=="behavior"))
		{
			if (nBehav > 1)
			{
				Report(c("Dependent actor variable ", net, ": ",
						netnames[net], "\n"), sep="", outf)
			}
			Report(c(" periods    actors:  down   up   constant  missing  ;",
					"   steps:   down    up  total\n"), sep="", outf)
			periodFromStart <- 0
			for (group in 1:nData)
			{
				j <- match(netnames[net], names(data[[group]]$depvars))
				if (is.na(j))
					stop("network names not consistent")
				depvar <- data[[group]]$depvars[[j]]
				atts <- attributes(depvar)
				for (i in 1:(atts$netdims[3] - 1))
				{
					mytab <- table(round(depvar[, 1, i + 1]) - 
						round(depvar[, 1, i]))
					numvals <- as.numeric(names(mytab))
					ups <- mytab[numvals > 0]
					downs <- mytab[numvals < 0]
					constants <- mytab[numvals == 0]
					stepsup <- sum(ups * numvals[numvals > 0])
					stepsdown <- sum(-1 * downs * numvals[numvals < 0])
					Report(c(format(i + periodFromStart, width=3),
							"  =>", format(i + 1 + periodFromStart, width=3),
							format(sum(downs), width=14),
							format(sum(ups), width=6),
							format(sum(constants), width=8),
							format(sum(is.na(depvar[, 1, i + 1]) |
									is.na(depvar[, 1, i])), width=10),
							format(stepsdown, width=21),
							format(stepsup, width=6),
							format(stepsup + stepsdown, width=6), "\n"),
						sep="", outf)
				}
				Report("\n", outf)
				periodFromStart <- periodFromStart + atts$netdims[3]
			}
			myeff <- effects[effects$name == netnames[net],]
			myobj <- myeff[myeff$shortName == "linear" &
				myeff$type == "eval",]
			Report(c("For this variable, the standard initial ",
					"behavioral tendency parameter is ",
					format(round(myobj$initialValue,4), nsmall=4,width=8),
					"\n"), sep="",
				outf)
		}

		Report("\n", outf)
	}
	
	##@initialContinuous internal printInitialDescription
	initialContinuous <- function()
	{
		Report("\n", outf)
		Heading(2, outf,"Dependent continuous actor variables:")
		Report("\n", outf)
		Heading(3, outf, "Changes")
		for (net in which(types=="continuous"))
		{
			Report(c("Dependent actor variable: ",
				netnames[net], "\n"), sep="", outf)
			Report(c(" periods    actors:  down   up   constant  missing  ;",
				" average:   down    up  total\n"), sep="", outf)
			periodFromStart <- 0
			for (group in 1:nData)
			{
				j <- match(netnames[net], names(data[[group]]$depvars))
				if (is.na(j))
					stop("network names not consistent")
				depvar <- data[[group]]$depvars[[j]]
				atts <- attributes(depvar)
				for (i in 1:(atts$netdims[3] - 1))
				{
					mytab <- table(depvar[, 1, i + 1] - depvar[, 1, i])
					numvals <- as.numeric(names(mytab))
					ups <- mytab[numvals > 0]
					downs <- mytab[numvals < 0]
					constants <- mytab[numvals == 0]
					averageup <- sum(ups * numvals[numvals > 0]) / length(ups)
					averagedown <- sum(downs * numvals[numvals < 0]) /
																  length(downs)
					averagechange <- sum(numvals) / length(numvals != 0) 
					Report(c(format(i + periodFromStart, width=3),
						"  =>", format(i + 1 + periodFromStart, width=3),
						format(sum(downs), width=14),
						format(sum(ups), width=6),
						format(sum(constants), width=8),
						format(sum(is.na(depvar[, 1, i + 1]) |
						is.na(depvar[, 1, i])), width=10),
						format(round(averagedown,2), width=21),
						format(round(averageup,2), width=6),
						format(round(averagechange,2), width=6), "\n"),
						sep="", outf)
					}
					Report("\n", outf)
					periodFromStart <- periodFromStart + atts$netdims[3]
				}
			}

		Report("\n", outf)
	}
	##### start of printInitialDescription function
	if (getDocumentation)
	{
		return(getInternals())
	}
	if (!inherits(data, "sienaGroup"))
	{
		nData <- 1
		data <- sienaGroupCreate(list(data), singleOK=TRUE)
	}
	else
	{
		nData <- length(data)
	}
	Report("\n\n", outf)
	Heading(1, outf, "Initial data description.")
	gpatts <- attributes(data)
	types <- gpatts$types
	netnames <- gpatts$netnames
	nobs <- gpatts$observations + length(data)
	nOneMode <- sum(types == "oneMode")
	nBehav <- sum(types == "behavior")
	nContinuous <- sum(types == "continuous")
	nBipartite <- sum(types == "bipartite")
	if (nOneMode  + nBipartite > 0)
	{
		initialNetworks()
	}
	if (nBehav > 0)
	{
		initialBehaviors()
	}
	if (nContinuous > 0)
	{
		initialContinuous()
	}
	Report(c("Initialisation of project <<", modelName,
			">> executed succesfully.\n"), sep="", outf)
}

Try the RSiena package in your browser

Any scripts or data that you put into this service are public.

RSiena documentation built on Nov. 2, 2023, 5:19 p.m.