HI: Time Inference of HI model
In QIU-Hongxiang-David/CAMer: Continuous Admixture Modeler

Description Usage Arguments Details Value Note See Also Examples

View source: R/CAMer.R

Estimate admixture time intervals/points for HI, CGF1(-I), CGF2(-I) and GA(-I) respectively for all Ld decay curves in a .rawld file.

1	HI(rawld, m1, T = 500L, LD.parallel = TRUE, LD.clusternum)

`rawld`	a string representing the path of the .rawld file or a data frame read from the .rawld file by `read.table`.
`m1`	the admixture proportion of population 1. If m2 is the admixing proportion of population 2, then m1+m2=1.
`T`	the most ancient generation to be searched. Defaults to 500.
`LD.parallel`	a logical expression indicating whether each LD decay curve should be computed parallely. Defaults to `TRUE`.
`LD.clusternum`	the number of clusters in parallel computation. See "Details".

LD.clusternum is used if LD.parallel=TRUE. If not specified, it is set to be the number of LD decay curves in the .rawld file.

The .rawld file should include exactly one column nameed "Distance" in Morgan, exactly one column named "Combined_LD", several columns named "Jack?" representing Jackknives where ? is a number and exactly one column named "Fitted" representing the fitted LD decay curve using the previous method. This function fits "Combined_LD" and all Jackknives using HI model. See singleHI for further details of fitting algorithm for each LD decay curve.

If the last entry of Distence in the .rawld file is greater than 10, a warning of unit will be given.

If the estimated time intervals/points cover T, a warning of too small T is given. The user should re-run the function with a larger T so that optimal time intervals/points can be reached.

Require parallel or snow package installed if LD.parallel=TRUE or single.parallel=TRUE. For newer versions of R (>=2.14.0), parallel is in R-core. If only snow is available, it is recommended to library it before using the parallel computing funcationality. When only snow is available, it will be require-d and hence the search path will be changed; if parallel is available, it will be used but the search path will not be changed. One may go to https://cran.r-project.org/src/contrib/Archive/snow/ to download and install older versions of snow if the R version is too old. If neither of the packages is available but LD.parallel=TRUE or single.parallel=TRUE, the function will compute sequentially with messages.

Be aware that when the computational cost is small (e.g. T=20L), using parallel computation for single LD decay curves can result in longer computation time.

an object of S3 class "CAM". A list CAM.list consisting of some basic information of function call, N objects of "CAM.single" class (where N is the number of LD decay curves in the .rawld file) calculated by singleHI, the fitted Ld decay curve fitted by the previous method (up to some truncation according to max.index) and a summary table containing the parameter estimates for each model and each curve, and diagnostic statistics msE and quasi-F. See singleCAM for more details of this kind of "CAM.songle" class.

There is a special method of print for this class.

Although the output is a "CAM" class object, it should NOT be passed to plot.CAM. Its summary table should NOT be passed to construct.CAM either.

When LD.parallel=TRUE, it is not recommended to terminate the execution of the function. If parallel package is available, it is said that setDefaultCluster from parallel can be used to remove the registered cluster, but real experiments do not support this; fortunately, these unused clusters will be removed automatically later, but with warnings. If only snow package is available, according to http://homepage.stat.uiowa.edu/~luke/R/cluster/cluster.html, "don't interrupt a snow computation". The ultimate method to close the unused clusters is probably to quit the R session.

singleHI, CAM

data(GA_I)
fit<-HI(GA_I,m1=0.3,T=150L)
fit

## Not run: 
#passing a file path to the argument `rawld=`
fit<-CAM("CGF_50.rawld",m1=0.3,T=150L)

## End(Not run)