mapStats: Calculate and plot survey statistics

Description

mapStats computes statistics and quantiles of survey variable(s), with optional class variables, and displays them on a color-coded map. It calls function calcStats which is also usable outside of mapStats.

Usage

mapStats(d, var, stat = c("mean", "quantile"), quantiles = c(0.5, 0.75),  wt.var = NULL, wt.label = TRUE, d.geo.var, by.var = NULL,  map.file, map.geo.var = d.geo.var, makeplot = TRUE, ngroups = 4,  separate = 1 , cell.min = 0, paletteName = "Reds", colorVec = NULL,  map.label = TRUE, map.label.names = map.geo.var, cex.label = 0.8, col.label = "black", titles = NULL, cex.title = 1, var.pretty = var, geo.pretty = map.geo.var, by.pretty = by.var, as.table = TRUE,  sp_layout.pars = list(), between = list(y = 1), horizontal.fill = TRUE, plotbyvar = ifelse( separate==1 & length(var)>1, FALSE, TRUE),  num.row = 1, num.col = 1, ...) 
calcStats(d, var, d.geo.var, stat = c("mean", "quantile"), quantiles=c(.5,.75),  by.var = NULL, wt.var = NULL, cell.min = 0)

Arguments

a data frame containing the variables to be analyzed.

var

a vector of character strings of the name of the variables that statistics will be calculated for. Multiple variables are allowed.

stat

a character vector of names of statistics to calculate. Valid names are "mean", "total", "quantile", "sd" (standard deviation), and "var" (variance). "Quantile" must be included for the quantiles specified to be calculated. Statistics are printed in the order given. For instance if stat = c("total","quantile","mean"), then the order will be total, then quantiles in order specified in argument quantiles, then the mean.

quantiles

a numeric vector of quantiles to be calculated for the variable var. The quantiles must be specified as decimals between 0 and 1. In order to be calculated, "quantile" must be specified as a statistic in the argument stat.

wt.var

a character string of the name of the variable to be used as sample weights in calculating statistics. The default is NULL, meaning unweighted statistics will be calculated.

wt.label

logical. Default is TRUE, in which case automatic titles will be followed by the string '(wtd.)' or '(unwtd.)' as appropriate, depending on whether weighted statistics were calculated. If FALSE no label will be added.

d.geo.var

a character string specifying the variable name in the dataset that is the geographic unit to calculate statistics by. When using calCstats outside of mapStats without a mapping application, d.geo.var would be the first class variable, and additional ones can be specified in by.var.

by.var

a vector of character strings specifying variable names in the dataset d to use as class variables. The order in which variables are specified will affect the order in which they are combined for calculations. The default is NULL, in which case statistics are calculated by each geographic unit (d.geo.var) only. The function will omit from analysis any class variables that have only one level over the entire dataset. However it is still possible that a given class variable will have only one value for one of the analysis variables if, say, multiple analysis variables are used.

map.file

an object of class SpatialPolygonsDataFrame on which the statistics will be plotted.

map.geo.var

a character string of the name of the geographic identifier in the data portion of map.file. The default is for it to be d.geo.var.The values of the geographic variables in the map and original dataset must be coded the same way for merging (i.e. the factor levels must be the same).

makeplot

logical. Default is TRUE; if FALSE, plots will not be drawn. This option can be used to calculate statistics without an available shapefile.

ngroups

a numeric vector of the number of levels for color plotting of variable statistics. If more than one number is specified, ngroups will be different in each plot. See details.

separate

numeric (or logical TRUE or FALSE for legacy cases). Accepted values are 0,1,2,3. This parameter controls how plot color breaks are calculated. If separate=0, all variables and statistic combinations have the same color breaks. If separate=3, each variable and statistic combination plot has a potentially different color break. If separate=1 (the default), then each statistic has a different color break, but this break is the same for the statistics acoss different variables. If separate=2, then each variable has a different color break, but this break is the same for all statistics of that variable. In the legacy case of this parameter, TRUE results in 1 and FALSE results in 0.

cell.min

numeric. Indicates the minimum number of observations in a cell combination of class variables of d.geo.var and by.var. If there are fewer than that, the statistic will be NA in that combination, and if plotted, the geographic unit will be white and not used in calculating the color key.

paletteName

a character vector containing names of color palettes for the RColorBrewer function brewer.pal. See details below for valid names. The default is to use these palettes for coloring, in which case ngroups will be restricted to between 3 and 9 levels, since there are at most 9 levels in RColorBrewer palettes. This is a good simple option. User-provided palettes can be used instead by specifying the argument colorVec to override this option. See details below.

colorVec

a list where each element is vector of ordered colors; they should be ordered from light to dark for a sequential palette. These override the use of RColorBrewer through the paletteName argument. See the demo for an example of using HCL sequential palettes from the colorspace package. Use of the colorVec argument will override a value provided for ngroups. See details below.

map.label

logical. Default is TRUE; if FALSE, names of the geographic regions will not be labeled on the map outputs.

map.label.names

a character string naming the vector from the map.file@data data.frame to use to label the map. The default is to use map.geo.var.

cex.label

numeric. Character expansion for the labels to be printed.

col.label

color of the label text to be printed. Default is black.

titles

a character string of length equal to the number of statistics to be plotted, in order. Replaces the default plot titles. The default is NULL, in which case the plot titles are automatically generated. See details below.

cex.title

numeric. Character expansion for the plot titles.

var.pretty

a character string used to name the analysis variable in the default plot titles. The default is to use var as the name in titles.

geo.pretty

a character string used to name the geographic variable in the default panel strip labels. The default is to use map.geo.var as the name labels.

by.pretty

a character string used to name the by-variables (optional class variables) in the default panel strip labels. The default is to use by.var as the name labels.

as.table

logical. Default is TRUE, meaning the panels will be displayed in ascending order of by.var (top to bottom).

sp_layout.pars

a list. This contains additional parameters to be plotted on each panel. See details section below and explanation of sp.layout in spplot. An example is provided in the demo file.

between

list. A lattice argument for parameters for spacing between panels.

horizontal.fill

logical. Default is TRUE, meaning that given the plot arrangement specified with num.row and num.col, plots will be plotted in order left to right then down. FALSE means they will be plotted going down first and then left to right. The user may need to use the optional lattice layout argument to control the layout of panels within a single plot to make sure the plots print with enough space, and par.strip.text to control the size of panel strip fonts. Examples are shown in the demo file.

plotbyvar

logical. If TRUE plots will be grouped by variable, otherwise by statistic.

num.row

numeric. To print multiple statistics on one page, indicate the number of rows for panel arrangement. Under the default, one statistic is printed per page.

num.col

numeric. To print multiple statistics on one page, indicate the number of columns for panel arrangement. Under the default, one statistic is printed per page.

...

Further arguments, usually lattice plot arguments.

Value

summary.stats: a list containing the calculated statistics matrices, as well as a frequency count matrix
variable: the name of the variable

Details

paletteName should contain one or more names of a sequential color palette in R from the RColorBrewer package. These are: Blues BuGn BuPu GnBu Greens Greys Oranges OrRd PuBu PuBuGn PuRd Purples RdPu Reds YlGn YlGnBu YlOrBr YlOrRd. The argument ngroups for this option should contain values between 3 and 9 since sequential color palettes have at most nine levels. The style argument from classIntervals can be included to change the method for calculating breaks (the default is by quantiles).

The default titles for the plots will be "(stat) of (variable) by (geography)", followed by either "(unwtd.)" or "(wtd.)", as appropriate. Using the wt.label argument controls the appearance of the weight label in the titles. Providing a value for the titles argument will override the default titles. This can be used, for instance, as shown below, to display percentages for a binary variable by calculating the mean of an indicator variable and specifying titles that indicate the percent is displayed.

If quantiles are 0 (minimum), 0.50 (median), or 1 (maximum), the statistics in the titles will be named "Minimum", "Median", and "Maximum" instead of "Q0", "Q50" or "Q100".

Parameter separate affects given values of colorVec, paletteName, and ngroups. Say there are 2 variables and 3 statistics (mean, median, and variance) to calculate for each. If any of the 3 parameters above are of length 1, (e.g. paletteName="Reds"), these will be used for all plots. If multiple values are given for any of the 3 parameters, these should be done with separate in mind. For instance, if separate=0 (same color breaks for all plots), then the first element of each of colorVec, paletteName, and ngroups will be used. If separate=1 (e.g. all of the means) have the same breaks, then the user may want to specify 3 different color palettes, for example. In this case, if ngroups=c(3,4,5,6), for instance, only the first 3 values will be used since there are only 3 statistics. If separate=2, then only up to 2 palettes, for instance, are used. If separate=3, then up to 6 values for each parameter will be used. See the demo file for examples.

The lattice layout argument can be used to control the placement of panels within a graph, especially if multiple plots are done on a page.

sp_layout.pars must itself be a list, even if its contents are lists also. This allows overplotting of more than one object. For instance, say you had a shapefile areas to be colored blue, and a vector of strings labels1 that had x-coordinates xplaces and y-coordinates yplaces to overlay on the plot. Create objects areas_overlay= list("sp.polygons", areas, fill="blue"), and labels_overlay= list("panel.text", labels1, xplaces, yplaces). Then set argument sp_layout.pars= list(areas_overlay, labels_overlay). Even if you only wanted to overlay with areas, you would still need to enclose it in another list, for example sp_layout.pars= list(areas_overlay). An example is shown in the demo file.

Examples

Run this code

#More complex examples with formatting are shown in the map_examples 
#demo for the package

#create synthetic survey dataset


state_codes <- c('AL','AK','AZ','AR','CA','CO','CT','DE','DC','FL','GA','HI','ID','IL',
                 'IN','IA','KS','KY','LA','ME','MD','MA','MI','MN','MS','MO','MT','NE',
                 'NV','NH','NJ','NM','NY','NC','ND','OH','OK','OR','PA','RI','SC','SD',
                 'TN','TX','UT','VT','VA','WA','WV','WI','WY')

surveydata <- data.frame(state=factor(rep(rep(state_codes, 
                         times=10), times=2)))
surveydata$year <- factor(rep(c(2009, 2010), each=nrow(surveydata)/2))
surveydata$obs_weight <- runif(n=nrow(surveydata), min=0.8, max=1.5)
surveydata$educ <- factor(sample(c("HS","BA","Grad"), replace=TRUE))

#two income distributions
surveydata$income <- 100000*rbeta(n=nrow(surveydata), 
                                  shape1=ifelse(surveydata$year==2009, 2, 1.5),
                                  shape2=ifelse(surveydata$year==2010, 10, 11))
surveydata$income_ge20k <- ifelse(surveydata$income >=20000, 100, 0)

#these state and year combinations will not be shaded if they are missing entirely
surveydata[ surveydata$state == "NV" & surveydata$year == 2009, 
            c("income","income_ge20k")] <- NA
surveydata[ surveydata$state == "OH" & surveydata$year == 2010, 
            c("income","income_ge20k")] <- NA


#load map shapefile
usMap <- readShapePoly(system.file("shapes/usMap.shp", package="mapStats")[1])


#Calculate weighted mean of variable income by state and year.  Display using red 
#sequential color palette with 4 groups.  In the titles, rename 'income'
#by 'household income'.     

mapStats(d=surveydata, var="income", d.geo.var="state", by.var="year",
         wt.var="obs_weight", map.file=usMap, stat="mean", ngroups=4, 
         paletteName="Reds", var.pretty="household income", map.label=TRUE)


#Calculate the weighted mean and 40th and 50th quantiles of the variable income
#by state, survey year, and education. Use three color palettes

## Not run:   
# mapStats(d=surveydata, var="income", d.geo.var="state", by.var=c("year","educ"),
#          wt.var="obs_weight", map.file=usMap, stat=c("mean","quantile"),     
#          quantiles=c(.4, .5), ngroups=6, paletteName=c("Reds","Greens","Blues"), 
#          var.pretty="household income", map.label=TRUE, num.col=1, num.row=2, 
#          layout=c(2,1), cex.label=.5)
# ## End(Not run)

#To calculate percentages of class variables, create an indicator variable, calculate
#its mean, and override the default titles to say you are calculating the percentage.
#Here we illustrate by calculating the percent of respondents by state that have income
#above $20,000.

## Not run: 
# mapStats(d=surveydata, var="income_ge20k", wt.var="obs_weight", 
#          map.file=usMap, d.geo.var="state", map.geo.var="STATE_ABBR", 
#          paletteName="Reds", stat=c("mean"), 
#          titles="Percent of respondents with income at least $20,000")
# ## End(Not run)


#calculating statistics using the functions outside of mapStats

## Not run: 
# calcStats(d=surveydata, var="income", stat="mean", 
#           d.geo.var="state", by.var=c("year", "educ"), wt.var="obs_weight")
# ## End(Not run)