no pcmax implementation in bd2kccd/r-causal

[kingfish777]

Greetings,
Please put a means to implement pcmax in r-causal. I am performing randomized sub-sampling of millions of rows/observations (and between 20 to 30K+ discrete binary variables) and I am going to do an alpha = .000005. I will be looking for stable edges across samples.
Thanks!
Scott

[kingfish777]

... which is strange because I see it here: https://github.com/bd2kccd/r-causal/blob/master/java/edu/cmu/tetrad/search/PcMax.java

[espinoj]

We've seen instances where the a recent R library is not loaded unless the old one is explicitly unloaded. You may want to try that. See this thread #23

…

On Feb 15, 2017, at 10:23 AM, Kingfish Threesevens ***@***.***> wrote: ... which is strange because I see it here: https://github.com/bd2kccd/r-causal/blob/master/java/edu/cmu/tetrad/search/PcMax.java — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub, or mute the thread.

[kingfish777]

Nope. Still not working. Detached, removed rcausal installations, library(devtools)/install_github'd, still no success (in seeing pcmax).
-S

[chirayukong]

Scott,

We don't currently support pcmax in the r-causal library yet. I'll put it in the next release. I'll notify you when it's done. In the mean time, you can use pcmax in the Tetrad application http://www.phil.cmu.edu/tetrad/.

Regards,

Kong

[kingfish777]

Hi Chiyaru, That's alright. I went ahead and wrote a wrapper for it and am debugging it (of course, in order to test it, NAMESPACE and INDEX in main folder have to be updated). I am testing it now.

…

--S ############ pcmax <- function(df, continuous = TRUE, depth = 3, #aggressivelyPreventCycles = FALSE, falseDiscoveryRate = FALSE, significance = 0.05, #maxPathLength = NULL, #useHeuristic = FALSE, verbose = FALSE, java.parameters = NULL, priorKnowledge = NULL) { params <- list(NULL) if(!is.null(java.parameters)){ options(java.parameters = java.parameters) params <- c(java.parameters = java.parameters) } # Data Frame to Independence Test indTest = NULL if(continuous){ tetradData <- loadContinuousData(df) indTest <- .jnew("edu/cmu/tetrad/search/IndTestFisherZ", tetradData, significance) }else{ tetradData <- loadDiscreteData(df) indTest <- .jnew("edu/cmu/tetrad/search/IndTestChiSquare", tetradData, significance) } indTest <- .jcast(indTest, "edu/cmu/tetrad/search/IndependenceTest") pcmax <- list() class(pcmax) <- "pcmax" pcmax$datasets <- deparse(substitute(df)) cat("Datasets:\n") cat(deparse(substitute(df)),"\n\n") # Initiate PC pcmax_instance <- .jnew("edu/cmu/tetrad/search/PcMax", indTest) .jcall(pcmax_instance, "V", "setDepth", as.integer(depth)) #.jcall(pcmax_instance, "V", "setMaxPathLength", as.integer(maxPathLength)) # .jcall(pcmax_instance, "V", "setUseHeuristic", useHeuristic) .jcall(pcmax_instance, "V", "setVerbose", verbose) if(!is.null(priorKnowledge)){ .jcall(pcmax_instance, "V", "setKnowledge", priorKnowledge) } params <- c(params, continuous = continuous) params <- c(params, depth = as.integer(depth)) # params <- c(params, aggressivelyPreventCycles = as.logical(aggressivelyPreventCycles)) # params <- c(params, falseDiscoveryRate = as.logical(falseDiscoveryRate)) params <- c(params, significance = significance) # params <- c(params, maxPathLength = as.integer(maxPathLength)) # params <- c(params, useHeuristic = as.logical(useHeuristic)) params <- c(params, verbose = as.logical(verbose)) if(!is.null(priorKnowledge)){ params <- c(params, prior = priorKnowledge) } pcmax$parameters <- params cat("Graph Parameters:\n") cat("continuous = ", continuous, "\n") cat("depth = ", as.integer(depth),"\n") # cat("aggressivelyPreventCycles = ", aggressivelyPreventCycles, "\n") # cat("falseDiscoveryRate = ", falseDiscoveryRate, "\n") cat("significance = ", as.integer(significance), "\n") cat("verbose = ", verbose, "\n") # cat("maxPathLength = ", as.integer(maxPathLength), "\n") # cat("useHeuristic = ", useHeuristic, "\n") # Search tetrad_graph <- .jcall(pcmax_instance, "Ledu/cmu/tetrad/graph/Graph;", "search") V <- extractTetradNodes(tetrad_graph) pcmax$nodes <- V # extract edges pcmax_edges <- extractTetradEdges(tetrad_graph) pcmax$edges <- pcmax_edges # convert output of PC into an R object (graphNEL) pcmax_graphNEL = tetradPattern2graphNEL(resultGraph = tetrad_graph, verbose = verbose) pcmax$graphNEL <- pcmax_graphNEL return(pcmax) }

________________________________ From: Chirayu (Kong) Wongchokprasitti <[email protected]> Sent: Wednesday, February 15, 2017 12:36 PM To: bd2kccd/r-causal Cc: Malec, Scott A; Author Subject: Re: [bd2kccd/r-causal] no pcmax implementation in bd2kccd/r-causal (#28) Scott, We don't currently support pcmax in the r-causal library yet. I'll put it in the next release. I'll notify you when it's done. In the mean time, you can use pcmax in the Tetrad application http://www.phil.cmu.edu/tetrad/<https://urldefense.proofpoint.com/v2/url?u=http-3A__www.phil.cmu.edu_tetrad_&d=DQMFaQ&c=6vgNTiRn9_pqCD9hKx9JgXN1VapJQ8JVoF8oWH1AgfQ&r=cJocORfYogykfmw40_PYbTR-_FlW5GZ15C5SG7IuxyE&m=sPSo1aFWOy5la4OwC_uzcEUJRPqfpsyqW303WMHeEyg&s=HpfbDpLoUfMWnYPsv8em3NqfB1ntzvXSQLkAvaM5N6o&e=>. Regards, Kong - You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub<https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_bd2kccd_r-2Dcausal_issues_28-23issuecomment-2D280098127&d=DQMFaQ&c=6vgNTiRn9_pqCD9hKx9JgXN1VapJQ8JVoF8oWH1AgfQ&r=cJocORfYogykfmw40_PYbTR-_FlW5GZ15C5SG7IuxyE&m=sPSo1aFWOy5la4OwC_uzcEUJRPqfpsyqW303WMHeEyg&s=QhV4WSxGIX1YW0Ksttgpax7tOpCFHGg1wG4rQ-ofI4Q&e=>, or mute the thread<https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_notifications_unsubscribe-2Dauth_ACh2SGWMlqcC-5FUMlXnfqDk3Lbr9xJziLks5rc0WbgaJpZM4MBWVv&d=DQMFaQ&c=6vgNTiRn9_pqCD9hKx9JgXN1VapJQ8JVoF8oWH1AgfQ&r=cJocORfYogykfmw40_PYbTR-_FlW5GZ15C5SG7IuxyE&m=sPSo1aFWOy5la4OwC_uzcEUJRPqfpsyqW303WMHeEyg&s=Z3EuNQLsAZGY61Yo5VprEehRN5GkOyPoFWvpJoXnwlg&e=>.

[chirayukong]

Thanks, Scott. :-) Anyway, the pcmax was added. If you have a chance, please check the new library out.