Analysing Data on Armis or Great Lakes using Rmd Files

Revision as of 17:06, 10 July 2023 by Davebridges (Talk | contribs) (Configuring R: added notes about creating R scripts)

Revision as of 17:06, 10 July 2023 by Davebridges (Talk | contribs) (Configuring R: added notes about creating R scripts)

Software

On OSX

On Windows

Access

At Michigan there are two clusters, one for human data (Armis) and one for non-PHI data (Great Lakes). Both of these are connected to a storage service called Turbo. These are available as part of the UM Research Computing Package.

Permissions


You can access the server either through the command line or through a remote desktop if you prefer a GUI. There a limited functionality to Armis2 here that can also be used (see these instructions).

If you submit a deidentified data request via https://datadirect.precisionhealth.umich.edu, your data will be sent as a 7zip file to the turbo storage in a folder named for your IRB. You will be prompted to enter a password to protect this file.

Accessing the Server - Command Line

On OSX open a terminal and enter ssh <UNIQNAME>@armis2.arc-ts.umich.edu replacing with your uniqname. Enter your level 1 password. Authenticate using Duo following the instructions. You will be placed in your home folder. Your data should be in /nfs/turbo/precision-health/DataDirect/<YOUR-IRB>. I find it convenient to make symlinks to quickly navigate to your data folder so first find your folder name

ls /nfs/turbo/precision-health/DataDirect/

locate the name of your folder <FOLDER_NAME>

create the symlink

ln -s /nfs/turbo/precision-health/DataDirect/<FOLDER_NAME> <LINK_NAME>

Now you can navigate from your home folder to your data folder by typing

cd <LINK_NAME>

Accessing the Server - Remote Desktop

Extracting Data

Submitting Scripts to the Server

Configuring R

Some R packages may need to be installed in your home folder. To do this go to your home folder cd ~/ and enter the following to load the R modules, enter an R shell, install the packages and exit out. You will only have to do this once to install the relevant R packages in your script

module load R
R
install.packages("PACKAGE_NAME")
exit()

Follow the prompts agreeing to make a personal library and using the Michigan mirror (71)

Creating a R script

I prefer to generate Rmd scripts on my own computer and then transfer them to the data folder to run

Transferring Results