optimStrata.RdWrapper function to call the different optimization functions: (i) optimizeStrata (method = "atomic"); (ii) optimizeStrata2 (method = "continuous"); (iii) optimizeStrataSpatial (method = "spatial"). For continuity reasons, these functions are still available to be used standalone.
optimStrata(method=c("atomic","continuous","spatial"), # common parameters framesamp, framecens=NULL, model=NULL, nStrata=NA, errors, alldomains=TRUE, dom=NULL, strcens=FALSE, minnumstr=2, iter=50, pops=20, mut_chance=NA, elitism_rate=0.2, suggestions=NULL, writeFiles=FALSE, showPlot=TRUE, parallel=TRUE, cores=NA, # parameters only for optimizeStrataSpatial fitting=NA, range=NA, kappa=NA)
| method | This parameter allows to choose the method to be applied in the optimization step: (i) optimizeStrata (method = "atomic"); (ii) optimizeStrata (method = "continuous"); (iii) optimizeStrata (method = "spatial") |
|---|---|
| errors | This is the (mandatory) dataframe containing the precision levels expressed in terms of maximum expected value of the Coefficients of Variation related to target variables of the survey. |
| framesamp | This is the dataframe containing the information related to the sampling frame. |
| framecens | This the dataframe containing the units to be selected in any case. It has same structure than "framesamp" dataframe. |
| nStrata | Indicates the number of strata to be obtained in the final solution. |
| model | In case the Y variables are not directly observed, but are estimated by means of other explicative variables, in order to compute the anticipated variance, information on models are given by a dataframe "model" with as many rows as the target variables. Each row contains the indication if the model is linear o loglinear, and the values of the model parameters beta, sig2, gamma (> 1 in case of heteroscedasticity). Default is NULL. |
| alldomains | Flag (TRUE/FALSE) to indicate if the optimization must be carried out on all domains (default is TRUE). If it is set to FALSE, then a value must be given to parameter 'dom'. |
| dom | Indicates the domain on which the optimization must be carried. It is an integer value that has to be internal to the interval (1 <--> number of domains). If 'alldomains' is set to TRUE, it is ignored. |
| strcens | Flag (TRUE/FALSE) to indicate if takeall strata do exist or not. Default is FALSE. |
| minnumstr | Indicates the minimum number of units that must be allocated in each stratum. Default is 2. |
| iter | Indicates the maximum number of iterations (= generations) of the genetic algorithm. Default is 50. |
| pops | The dimension of each generations in terms of individuals. Default is 20. |
| mut_chance | Mutation chance: for each new individual, the probability to change each single chromosome, i.e. one bit of the solution vector. High values of this parameter allow a deeper exploration of the solution space, but a slower convergence, while low values permit a faster convergence, but the final solution can be distant from the optimal one. Default is NA, in correspondence of which it is computed as 1/(vars+1) where vars is the length of elements in the solution. |
| elitism_rate | This parameter indicates the rate of better solutions that must be preserved from one generation to another. Default is 0.2 (20 |
| suggestions | Optional parameter for genetic algorithm that indicates a suggested solution to be introduced in the initial population. The most convenient is the one found by the function "KmeanSolution". Default is NULL. |
| writeFiles | Indicates if the various dataframes and plots produced during the execution have to be written in the working directory. Default is FALSE. |
| showPlot | Indicates if the plot showing the trend in the value of the objective function has to be shown or not. In parallel = TRUE, this defaults to FALSE Default is TRUE. |
| parallel | Should the analysis be run in parallel. Default is TRUE. |
| cores | If the analysis is run in parallel, how many cores should be used. If not specified n-1 of total available cores are used OR if number of domains < (n-1) cores, then number of cores equal to number of domains are used. |
| fitting | Fitting of the model(s) (in terms of R squared). It is a vector with as many elements as the number of target variables Y. |
| range | Maximum range for spatial autocorrelation. It is a vector with as many elements as the number of target variables Y. |
| kappa | Factor used in evaluating spatial autocorrelation. |
List containing (1) the vector of the solution, (2) the optimal aggregated strata, (3) the total sampling frame with the label of aggregated strata
Giulio Barcaroli
if (FALSE) { library(SamplingStrata) ############################ # Example of "atomic" method ############################ data(swissmunicipalities) swissmunicipalities$id <- c(1:nrow(swissmunicipalities)) frame <- buildFrameDF(df = swissmunicipalities, id = "id", domainvalue = "REG", X = c("POPTOT","HApoly"), Y = c("Surfacesbois", "Airind")) ndom <- length(unique(frame$domainvalue)) cv <- as.data.frame(list(DOM = rep("DOM1",ndom), CV1 = rep(0.1,ndom), CV2 = rep(0.1,ndom), domainvalue = c(1:ndom))) strata <- buildStrataDF(frame) kmean <- KmeansSolution(strata,cv,maxclusters=30) nstrat <- tapply(kmean$suggestions, kmean$domainvalue, FUN=function(x) length(unique(x))) solution <- optimStrata(method ="atomic", framesamp = frame, errors = cv, nStrata = nstrat, suggestions = kmean, iter = 50, pops = 10) outstrata <- solution$aggr_strata framenew <- solution$framenew s <- selectSample(framenew, outstrata) ################################ # Example of "continuous" method ################################ kmean <- KmeansSolution2(frame = frame, errors = cv, maxclusters = 10) nstrat <- tapply(kmean$suggestions, kmean$domainvalue, FUN=function(x) length(unique(x))) sugg <- prepareSuggestion(kmean = kmean, frame = frame, nstrat = nstrat) solution <- optimStrata(method = "continuous", framesamp = frame, errors = cv, nStrata = nstrat, iter = 50, pops = 10, suggestions = sugg) framenew <- solution$framenew outstrata <- solution$aggr_strata s <- selectSample(framenew,outstrata) ############################# # Example of "spatial" method ############################# library(sp) data("meuse") data("meuse.grid") meuse.grid$id <- c(1:nrow(meuse.grid)) coordinates(meuse) <- c('x','y') coordinates(meuse.grid) <- c('x','y') library(gstat) library(automap) v <- variogram(lead ~ dist + soil, data = meuse) fit.vgm.lead <- autofitVariogram(lead ~ dist + soil, meuse, model = "Exp") plot(v, fit.vgm.lead$var_model) lead.kr <- krige(lead ~ dist + soil, meuse, meuse.grid, model = fit.vgm.lead$var_model) lead.pred <- ifelse(lead.kr[1]$var1.pred < 0,0, lead.kr[1]$var1.pred) lead.var <- ifelse(lead.kr[2]$var1.var < 0, 0, lead.kr[2]$var1.var) df <- as.data.frame(list(dom = rep(1,nrow(meuse.grid)), lead.pred = lead.pred, lead.var = lead.var, lon = meuse.grid$x, lat = meuse.grid$y, id = c(1:nrow(meuse.grid)))) frame <-buildFrameSpatial(df = df, id = "id", X = c("lead.pred"), Y = c("lead.pred"), variance = c("lead.var"), lon = "lon", lat = "lat", domainvalue = "dom") cv <- as.data.frame(list(DOM = rep("DOM1",1), CV1 = rep(0.05,1), domainvalue = c(1:1) )) solution <- optimStrata(method = "spatial", errors = cv, framesamp = frame, iter = 25, pops = 10, nStrata = 5, fitting = 1, kappa = 1, range = fit.vgm.lead$var_model$range[2]) framenew <- solution$framenew outstrata <- solution$aggr_strata frameres <- SpatialPixelsDataFrame(points = framenew[c("LON","LAT")], data = framenew) frameres$LABEL <- as.factor(frameres$LABEL) spplot(frameres,c("LABEL"), col.regions=bpy.colors(5)) s <- selectSample(framenew,outstrata) }