--- title: Fitting GPR and generation of features for new test set with 100k light curves author: Julian Faraway output: html_document: toc: true theme: cosmo --- ```{r global_options, include=FALSE} library(knitr) opts_chunk$set(comment=NA, fig.path='/tmp/Figs/', warning=FALSE, message=FALSE) ``` ```{r echo=FALSE} paste("Created:",date()) ``` Load in the new data. Load code from previous work. ```{r} load("lc100k.rda") row.names(lcdbi) <- lcdbi$id lcdbi$id <- NULL lcdbi$type <- factor(lcdbi$type) source("funcs.R") GPR <- TRUE ``` Set the kernel type to squared exponential and set the inverse width: ```{r} kerntype <- "exponential" wvec <- 2e-4 ``` ```{r} summary(lcdbi) ``` We see no short light curves. Investigate the range of observation. Compute the date of the first observation: ```{r} fd <- sapply(lcdb, function(x) min(x$mjd)) min(fd) ``` We see that 53464 is indeed the start date. ```{r} robv <- sapply(lcdb, function(x) max(x$mjd) - min(x$mjd)) max(robv) ``` Seems we have substantially shorter range of observation in this dataset than the 2764 used previously. Sort the lightcurves to be monotonely increasing in date and create the measurement grouping variable: ```{r} for(i in 1:length(lcdb)){ lc <- lcdb[[i]] ii <- order(lc$mjd) lc <- lc[ii,] lc$measgrp <- factor(c(0,cumsum(ifelse(diff(lc$mjd) < 1, 0, 1)))) lcdb[[i]] <- lc } ``` ## Detection Limit For each light curve, we compute - proportion of observations with magnitudes greater than 20.5, 21 and 21.5 - the maximum magnitude ```{r} dlmat <- matrix(NA,length(lcdb),4) for(i in 1:length(lcdb)){ lc <- lcdb[[i]] nobs <- nrow(lc) dl1 <- sum(lc$mag > 20.5)/nobs dl2 <- sum(lc$mag > 21)/nobs dl3 <- sum(lc$mag > 21.5)/nobs dl4 <- max(lc$mag) dlmat[i,] <- c(dl1, dl2, dl3, dl4) } colnames(dlmat) <- c("d20.5","d21","d21.5","min") dlmat <- data.frame(dlmat) summary(dlmat) ``` Looks like 20.5 is a suitable detection limit here. Now ready to compute the measure. ```{r} firstdate <- 53464 daterange <- 2222 detection.limit <- 20.5 lcdbi$id <- row.names(lcdbi) ``` ```{r child="meascomp.Rmd",eval=TRUE} ``` ```{r} cmdb$wander <- NULL cmdb$moveloc <- NULL summary(cmdb) ``` ```{r} for(i in 1:(ncol(cmdb)-2)){ plot(cmdb[,i] ~ cmdb$type, main=names(cmdb)[i]) } ``` ```{r} cmdbig <- cmdb save(cmdbig,file="featbig.rda") ```