Federico A. Bugni and Joel L. Horowitz, "Permutation Tests for
Equality of Distributions of Functional Data", Journal of Applied
Econometrics, Vol. 36, No. 7, 2021, pp. 861-877.

The data used in this paper were produced by the smart metering
consumer behavior trial (CBT) for gas conducted by the Commission for
Energy Regulation (CER) of Ireland. We obtained the data from the
Irish Social Science Data Archive (ISSDA). These data are subject to a
confidentiality agreement. To access the data, please contact the
ISSDA (http:www.ucd.ie/issda).

The data include several files with documentation. Our application
combined the information from the following files:

-- ``residential allocations.xls'': this Excel file indicates the
household treatment assignment. This file has two columns labeled
``ID'' and ``Assignment''. Column ID indicates the household number,
and column ``Assignment'' indicates the treatment assignment.

-- ``GasDataWeekX'', where X = 0,1,...,77: each one of these ASCII
files provides the household gas consumption during a specific week,
measured every half hour. These 78 files span the period going from
December 8th, 2009 until May 30th, 2011. Each one of these files has
three columns labeled ``ID'', ``DT'', and ``Usage''. Column ID
indicates the household number (which coincides with ``residential
allocations.xls''), ``DT'' is a five-digit code that indicates day and
time of measurement, and ``Usage'' is gas consumption measured in
kiloWatt.

-- The five-digit code for ``DT'' is as follows. First, digits 1-3 of
``DT'' indicate the day, with 001 indicating January 1st, 2009, 002
indicating January 2nd, 2009, and so on until 880, which indicates May
30th, 2011. Second, digits 4-5 of ``DT'' indicate the 30-minute
interval within the day, with 01 indicating the first half-hour of the
day (i.e., 00:00:00-00:29:59), 02 indicating the second half-hour of
the day (i.e., 00:30:00-00:59:59), and so on until 48 indicating the
last half-hour of the day (i.e., 23:30:00-23:29:59). The CBT began on
the first half-hour of June 1st, 2010, which corresponds to DT=51701.

The raw dataset has information on 1,576 households, but 84 of these
households have missing gas consumption information. Given that the
proportion of households with missing data is small and a formal
analysis of the missing data problem is out of our scope, we decided
to eliminate the households with missing information. This generated
our final dataset with 1,492 households. For each one of these
households, we observe their treatment assignment (from ``residential
allocations.xls'') and their half-hour gas consumption between June
1st, 2010 through December 31st, 2010 (from ``GasDataWeekX'' with X =
26,...,56). 

We also include two additional Matlab m-files:

-- ``Gas_Cleaning_File.m'': this m-file reads the original ISSDA data,
cleans it according to the description in the previous paragraph, and
generates Matlab data that is used to implement our test. To use this
program, place it in the same directory as the ISSDA data.

-- ``Gas_MC_and_Figures.m'': this m-file creates the one-month data
required for the Monte Carlo simulations and generates the figures
shown in the paper. This program should be run after
``Gas_Cleaning_File.m'' has been completed.

The two Matlab files are in DOS format. They are zipped in the file
bh-files.zip.