```{r echo=FALSE, message=FALSE, warnings=FALSE} opts_chunk$set(echo = FALSE, messages = FALSE, fig.height = 6, fig.width = 6) ``` # Austrian national guest survey: AA versus PAA ```{r results='hide', message=FALSE, warning=FALSE, error=FALSE} source("helpers.R") ``` This document reproduces the comparison of the probabilistic and the classical archetypes in the example "Austrian national guest survey" presented in the paper "[Probabilistic archetypal analysis](http://arxiv.org/abs/1312.7604)" by Seth and Eugster (2014). The [Rmd file](GSAW97-report-aa.Rmd) lists the corresponding [R](http://www.r-project.org) code. ```{r results='asis', message=FALSE, warning=FALSE} cat("* Read tourist data from file `GSAW97_bin.csv`\n") GSAW97bin <- read.csv2("data_GSAW97/GSAW97_bin.csv") cat("* Load precomputed PAA solution from file `GSAW97_paa.mat`\n") paa <- load_paa("data_GSAW97/GSAW97_paa.mat", GSAW97bin) cat("* Load precomputed AA solution from file `GSAW97_aa.Rds`\n") aa <- readRDS("data_GSAW97/GSAW97_aa.Rds") cat("* Match AA solution with PAA solution\n") match <- greedy_match(aa$archetypes, t(paa$Z)) match <- match[c(1, 3, 5, 4, 6, 2), ] ``` #### Probabilistic archetypes: ```{r} thepaa <- paa$Z colnames(thepaa) <- sprintf("PAA%s", 1:6) round(thepaa[, match[, "PAA"]], 1) ``` #### Classical archetypes: The classical archetypes were computed with the [archetypes](cran.r-project.org/package=archetypes) package. The presented solution is the best solution out of twenty repetitions. ```{r} theaa <- t(aa$archetypes) colnames(theaa) <- sprintf(" AA%s", 1:6) round(theaa[, match[, "AA"]], 1) ```