Stata

Stata is data analysis and statistical software¹.

Installed Versions

Stata 19 and Stata 18 are installed. We have a 48-core (since there are 48-cores per node on Picotte), ten seat license.

You can use Stata by running:

module load stata/mp48/19

or

module load stata/mp48/18

Depending which version you want to use.

Documentation

PDF documentation is available on Picotte. They are in the directory ${STATADIR}/docs/ available once the modulefile is loaded.

You can also use Stata's online documentation

Personal Setup

Your own ADO files go in ~/ado/personal/².

Running

Command Line Version

To run the command line version:

[juser@picotte001 ~]$ module load stata/mp48/19
[juser@picotte001 ~]$ stata

  ___  ____  ____  ____  ____ ®
 /__    /   ____/   /   ____/      18.0
___/   /   /___/   /   /___/       BE—Basic Edition

 Statistics and Data Science       Copyright 1985-2023 StataCorp LLC
                                   StataCorp
                                   4905 Lakeway Drive
                                   College Station, Texas 77845 USA
                                   800-STATA-PC        https://www.stata.com
                                   979-696-4600        stata@stata.com

Stata license: 10-user network, expiring 24 Feb 2026
Serial number: 501809306532
  Licensed to: Drexel University
               Philadelphia

Notes:
      1. Unicode is supported; see help unicode_advice.

.

Graphical User Interface (GUI) Version

To run the GUI version, you must have an X11 server installed on your computer. See:

The command to use is xstata.

640px

Note that if you do this, you will be running Stata on the login node, which is a shared resource where multiple people may be logged in and doing work simultaneously.

Submitting Jobs

For long-running computations, you will want to write a job script to be submitted as a job on the cluster.

Please see more detail in Writing Slurm Job Scripts

Requesting License

Each job should request one license. The license on Picotte is limited to no more than 10 simultaneous uses.

#SBATCH --licenses=stata48:1

Stata/MP

Stata/MP, which runs multithreaded, is also available. It is provided by the command stata-mp. There are two editions of Stata/MP on Picotte: one licensed for 4 cores (unlimited number of seats), and one licensed for 48 cores (limit of ten seats).

N.B. more cores (threads) does not guarantee better performance. It can frequently be the reverse.

NOTE: you must do "set processors NN" in your .do file to be the exact number of slots requested by the job. The number of slots (each slot is one processor core) is requested by the lines:

#SBATCH --nodes=1
#SBATCH --ntasks=1
#SBATCH --cpus-per-task=48
#SBATCH --license=stata48:1

module load stata/mp48/19

In this example, the number of CPU cores is 48.

You may also read the environment variable SLURM_CPUS_PER_TASK to use in the "set processors NN" command in your Stata .do file:

local p : env SLURM_CPUS_PER_TASK
set processors $p

SLURM_CPUS_PER_TASK is set by Slurm, the job scheduler, to be the value requested by "--cpus-per-task".

NOTE ON PERFORMANCE: More does not necessarily mean faster. Some functions/routines may be parallelizable, others may not be. You will need to benchmark your specific computation to find the optimal number of CPU cores to use in the computation.

Example Job for Picotte (Slurm)

This is the Stata script to be run -- named testing.do:

// test computation - testing.do
clear*
set rmsg on
set obs 100000
local p : env SLURM_CPUS_PER_TASK
set processors $p

forval n = 1/5 {
    g i`n' = runiform()
}
g dv = rbinomial(1,.3)
memory

qui logit dv i*

qui xtmixed dv i*

*with bootstrap:
qui bs, reps(2000): logit dv i*

This is the job script -- named teststata.sh:

#!/bin/bash
#SBATCH --nodes=1
#SBATCH --ntasks=1
#SBATCH --cpus-per-task=4
#SBATCH --mem=128G
#SBATCH --account=urcfadmprj
#SBATCH --time=4:00:00
#SBATCH --license=stata48:1

module load stata/mp48/19

# set the Stata temporary directory to the job-specific temporary directory
# this directory will be automatically deleted at the end of the job
export STATATMP=$TMP

stata-mp -b do testing.do

To submit the job:

[juser@picotte001]$ sbatch teststata.sh

NB you may see a warning/error message in the Slurm output file; this can be safely ignored.

stata-mp: /lib64/libtinfo.so.5: no version information available (required by stata-mp)

Outputs

Stata, by default, produces a log file named after the .do file. So, running the Stata DO script something.do produces the log something.log

If the same DO script is run multiple times, later runs will overwrite the log from earlier runs.