SUMM: Non-reproducible file references

[jbrockmendel]

Most of these are paths to R or Stata scripts that are used to create a results files. These need to be pointed towards relative paths in the repo in order for these to be reproducible.

• koul_and_mc.py calls np.genfromtxt('/home/justin/rverify.csv', ...)
• iolib/tests/gen_dates.do insheet using "/home/skipper/statsmodels/statsmodels-skipper/statsmodels/iolib/tests/stata_dates.csv"
• sandbox/examples/example_sysreg.py data3 = np.genfromtxt('/home/skipper/school/MetricsII/Greene 150 TableF5-1.txt', names=True)
• sandbox/tests/GreeneEx15_1.s dta <- read.table('/home/skipper/school/MetricsII/Greene\ TableF5-1.txt', header = TRUE)
• stats/tests/results/anova.R source('/home/skipper/statsmodels/statsmodels/tools/topy.R')
• tsa/tests/arima.do insheet using "/home/skipper/statsmodels/statsmodels-skipper/statsmodels/datasets/macrodata/macrodata.csv" (and a couple others)
• tsa/tests/arima112.do insheet using "/home/skipper/statsmodels/statsmodels-skipper/statsmodels/datasets/macrodata/macrodata.csv" (and a couple others)
• tsa/tests/arima211.do insheet using "/home/skipper/statsmodels/statsmodels-skipper/statsmodels/datasets/macrodata/macrodata.csv" (and a couple others)
• tsa/tests/results/arima.R dta <- read.csv("/home/skipper/statsmodels/statsmodels-skipper/statsmodels/datasets/macrodata/macrodata.csv")
• tsa/tests/results/arima_forecast.inp open /home/skipper/statsmodels/statsmodels/statsmodels/datasets/macrodata/macrodata.csv (one other)
• tsa/tests/results/arma_forecast.inp open /home/skipper/statsmodels/statsmodels/statsmodels/tsa/tests/results/y_arma_data.csv
• tsa/tests/results/corrgram.do insheet using "/home/skipper/statsmodels/statsmodels-skipper/statsmodels/datasets/macrodata/macrodata.csv", double clear (one other)
• tsa/vector_ar/tests/svar.do insheet using "/home/skipper/statsmodels/statsmodels-skipper/scikits/statsmodels/datasets/macrodata/macrodata.csv", double clear
• tsa/vector_ar/tests/svartest.R data <- read.csv("C:\\statsmodels\\statsmodels-bartbkr\\scikits\\statsmodels\\datasets\\macrodata\\macrodata.csv") (and 2 more commented-out)
• tsa/vector_ar/tests/var.R data <- read.csv('/home/wesm/code/statsmodels/scikits/statsmodels/datasets/macrodata/macrodata.csv')
• stats/libqsturng/CH.r setwd('D:\\USERS\\roger\\programming\\python\\development\\qsturng')
• gam/tests/results/results_mpg_bs_poisson.r source("M:\\josef_new\\eclipse_ws\\statsmodels\\statsmodels_py34_pr\\tools\\R2nparray\\R\\R2nparray.R") (@josef-pkt I've seen R2nparray used elsewhere with just library(R2nparray); any chance that would work here?)
• tsa/statespace/tests/results/test_var.R dta <- read.dta13('~/projects/statsmodels/statsmodels/tsa/tests/results/lutkepohl2.dta') (one other)
• tsa/statespace/tests/results/test_exact_diffuse_filtering_multivariate.R setwd('~/projects/statsmodels-0.9/statsmodels/')
• tsa/statespace/tests/results/test_exact_diffuse_filtering.R setwd('~/projects/statsmodels-0.9/statsmodels/')
• regression/tests/results/test_rls.R macrodata <- read.csv('/Users/fulton/projects/statsmodels/statsmodels/datasets/macrodata/macrodata.csv') (one other)
• regression/tests/results/test_rls.do insheet using /Users/fulton/projects/statsmodels/statsmodels/datasets/macrodata/macrodata.csv, clear
• compare_arma.py references r'C:\Josef\eclipsegworkspace\statsmodels-josef-experimental-gsoc\scikits\statsmodels\tsa\y_arma22.txt'; that file doesn't appear to exist anymore.

[josef-pkt]

Doesn't work or not worth the trouble.
Every one who wants to run the Stata or R files can replace the path.

I never tried to figure out if there would be a way to find the statsmodels file location from within Stata.
In any case, it is not worth the effort.
Same for R

(I'm running my Stata files from a stataworks directory and then copy the relevant files into statsmodels, and similarly for R.)

[jbrockmendel]

I care quite a bit about reproducibility, so do judge it worth the effort.

Some of these can just be updated to point towards local paths. For others the original file is not in the repo, and they need to be tracked down.

[josef-pkt]

Have you written a doc notebook yet? Or some additional unit tests using R or Stata?

We are not going to impose extra work on contributors that do provide R/Stata verified unit tests.

[jbrockmendel]

Any thoughts on the R2nparray question? Also #5719 should be answerable in one sentence.