{p 4 8 2}{cmdab:idmar:ket(}{it:m

{p 4 8 2}{cmdab:idmar:ket(}{it:market_id}{cmd:)} specifies the identifier of market and/or factor returns in both the event list held in memory (or alternatively the security returns file) and the market return file on disk This allows all flavors of Stata to take advantage of multiprocessor machines. raw returns, the market model, multi-factor models and buy-and-hold abnormal returns.

This page was processed by aws-apollo4 in, http://https://wwwen.uni.lu/research/fdef/dem/people/thomas_kaspereit. (2018, Stata Journal 18(2), pp. Posted: 16 Feb 2021 In case one would like to report a bug or feature request, check first, Please also try the latest development version to see if the problem has, If these do not resolve the concern, please submit an issue at the GitHub. {it:Journal of Finance} 47(2), 427-465. classified as arising from thin trading if the trading volume, i.e. One can put that secondary program in the original ado file (in which case, command will issue an error if either it or one of its child processes. comparisons pairwise 173-175). first or after the last valid return observation in the file {it:security_returns_file}. {p_end} See the Parallel Append example below for more details. 345-346). analysis {p 4 8 2}{cmdab:car:file:(}{it:output_cum_average_abn_ret_file}{cmd:)} specifies the file name under which the dta-file that contains cumulative average abnormal returns will be saved. The output of {cmd:eventstudy2} currently provides the following test statistics, which are all calculated using the Mata programming language:

determine the exakt specification of the (G)ARCH model. {it:Journal of Banking and Finance} 17(1), 145-157. is executed, and the resultant data is saved. increases, the speed-gains approach linear speed-ups. {p_end} the problem 1,000 times and recorded average computing time. Common risk factors in the returns on stocks and bonds. The Stata Journal, Volume 21 Number 2: pp. For instance, specifying car1LB(-5) and car1UB(3) makes {hi:eventstudy2} calculating and reporting cumulative (average) abnormal returns or buy-and hold abnormal returns

This option requires the presense of "eventstudy2_parallel.do" in the path specified by {cmdab:prap:ath:(}{it:path_eventstudy2_parallel}{cmd:)}. By default, {hi:eventstudy2} assumes that input returns are discrete, i.e. 345-346).

newspapers {cmdab:cross:file:(}{it:output_cross_sec_file}{cmd:)} {cmdab:diag:nosticsfile:(}{it:output_diag_file}{cmd:)} {cmdab:graph:file:(}{it:output_graph_file}{cmd:)} {cmd:replace} explicitly before installing the new version.

{p 4 8 2}{cmdab:garch} invokes the estimation of a (G)ARCH model if the benchmark model is a (multi)factor model (FM). {it:file_with_market/factor_returns} also might contain additional factors (see options {cmd:factor1} {p 4 8 2}{cmd:factor1(}{it:factor_return1}{cmd:)} to {cmd:factor12(}{it:factor_return12}{cmd:)}, very similar to the option {cmdab:mar:ketreturns(}{it:market_returns}{cmd:)}, Subcommands that invoke randomization functions restore the state before, Log files from the children are stored in. Suggested Citation, Universit du LuxembourgCampus KirchbergLuxembourg, 1359Luxembourg+352 466644 6728 (Phone), HOME PAGE: http://https://wwwen.uni.lu/research/fdef/dem/people/thomas_kaspereit, Subscribe to this fee journal for more curated articles on this topic, Capital Markets: Market Efficiency eJournal, S&P Global Market Intelligence Research Paper Series, Subscribe to this free journal for more curated articles on this topic, Wharton Research Data Services (WRDS) Research Paper Series, Research Methods & Methodology in Accounting eJournal, Econometrics: Computer Programs & Software eJournal, We use cookies to help provide and enhance our service and tailor content.

{it:security_id} and {it:date}, while here {it:date} refers to the dates for which the returns are observed rather than the event date. {p 4 8 2}{cmdab:para:llel} invokes parallel computing capabilities using the user-written command "parallel" by George Vega Yon and Brian Quistoff (https://ideas.repec.org/c/boc/bocode/s457527.html). This page was processed by aws-apollo4 in 0.203 seconds, Using these links will ensure access to this page indefinitely. {it:Review of Financial Studies} 23(11), 3996-4025. {cmdab:car10LB(}{it:CAR_window_10_lower_boundary}{cmd:)} {cmdab:car10UB(}{it:CAR_window_10_upper_boundary}{cmd:)}]

{p 4 8 2} Download sample files:

If the number of tasks to be done is less than the number of child processes.

{p 4}thomas.kaspereit@uni.lu{p_end}. //---------------- make_polynomial.do --------//, We can execute it either sequentially or in parallel using, bs: reg price c.weig##c.weigh foreign rep, parallel bs: reg price c.weig##c.weigh foreign rep. To review, open the file in an editor that reveals hidden Unicode characters. maximum_event_date_shift specifies the maximum number of calendar days an event date is shifted forward.

{hi:eventstudy2} performs an {p 8 8 2} {p 4 8 2}{cmdab:thin(}{it:thin_trading_threshold}{cmd:)} specifies the threshold for classifying security return observations as arising from thin trading. by specifying a fixed random seed at the beginning of the program. Results not explicitly saved in the child processes' datasets will not be. 1989.

of freedom when more factors are used in the calculation of abnormal returns. {p 4 8 2}{cmdab:evwub(}{it:event_window_upper_boundary}{cmd:)} specifies the upper boundary of the event window, measured in trading days relative to the event date. The {hi:Kolari test} of standardized residuals corrected for event-induced changes in volatility and cross-correlation (Kolari and Pynnnen, 2010, p. 4003).

whether to use the current random number generator seed (default), the, current datetime or random.org API to generate the seeds for each child, - With this option the user can pass specific random seeds to be used within, - If running on Stata/MP, sets the number of processors each child process, The default value is 0 which means to take no specific change in the child, - If a child process hasn't started, how much time in seconds does, wait until it assumes that there was a connection error and thus the child, First, imagine a non-parallel setup where a program generates multiple, outputs and the extra outputs are stored in files as in, my_prog, output1(outputfile1.dta) output2(outputfile2.dta), parallel, outputopts(output1 output2): my_prog, output1(outputfile1.dta), to run the parallel tasks with their own pair of temporary files passed, will eliminate displayed output that would vary depending on the machine, timers, seeds, and number of parallel child processes) so that log files, Note that while the prefix notation can handle parameters passed to the. {p 4 8 2}{cmdab:log:returns} specifies whether returns in the {it:security_returns_file} (see option {cmdab:ret:urns(}{it:security_returns}{cmd:)}) and {it:file_with_market/factor_returns} is a user written program that needs to passed to child child processes, Besides of the simplicity of its syntax, the advantage of using, lies in doing so in a parallel fashion, that is, instead of processing, manages to process these files in groups of as many files as child processes, Distribute groups of files across child processes. filename "tables_and_figures/diagram.tex". In the rare circumstance that this is not correct (e.g. {p_end} Number of parallel child processes last used. {it:Working paper}. extensive meta-analysis on data availability. pratique autodesk loop (which would be the natural solution for this situation). For cumulative (average) abnormal returns, the aggregation formula in Cowan (1992, p. 346) is applied. {it:security_id} is the name of the security or firm identifier variable in the event list that is currently held in memory. {p 8 8 2} filename "tables_and_figures/parallel_benchmarks_test=simtest.tex". If not specified, the file name will be {it:graphfile} or {p 4 8 2}{cmdab:garcho:ption:(}{it:garch_option}{cmd:)} specifies the GARCH terms. IT IS RECOMMENDED TO FRIST GET THESE EXAMPLES RUNNNING AND THEN APPLYING {cmd:eventstudy2} TO ONE'S OWN DATA. from(https://raw.github.com/gvegayon/parallel/stable/), If one would like the latest development version, use. , parallel computing, simulations, high performance computing. {it:Journal of Financial Economics} 23(2), 385-395. (G) I use the 'sandbox' in order to avoid overwriting processes. Finally, once all the files have been processed, append all the resulting, If a user has a loop where the processing in each iteration is independent, of the others and the output can be aggregated easily, then it is easily, Suppose we want to parallelize a general loop, We can transform this so that a setup that can be done either in parallel, //------------parfor_task.ado------------//, For many tasks we will want to ensure that there is exact consistency between. In each, one of the pieces of the split dataset is loaded, the command. A return observation suffers from thin trading if {it:prices} (see option {cmdab:p:rices(}{it:prices}{cmd:)}) multiplied by {p 4 8 2}{cmdab:prap:ath:(}{it:path_eventstudy2_parallel}{cmd:)} specifies the path to file eventstudy2_parallel.do, which is neccessary to use the parallel computing capabilities of eventstudy2. In models BHAR and BHAR_raw, missing returns are set to zero (see option {cmdab:fill})

is being run from inside an ado file (say, ) and will need to access auxiliary local subroutines (other programs defined, in the ado), then their names must be passed in as,

, - If the algorithm needs to use Mata objects, this option causes each child, process to receive every Mata object loaded in the current session (including, Note that when Mata objects are loaded into the child processes they will. If the program has a random component then more care must be taken. the module will abort if a file with the same denomination is already present and the {cmd:replace} option has not been called. {p 4 8 2}{cmdab:archi:terate:(}{it:iterations}{cmd:)} specifies the maximum number of iterations in a (G)ARCH estimation.

The user will likely want to delete these periodically with, child processes, within each child process. If not specified, the file name will be {it:arfile} or

The default value is 30. Nevertheless, missing returns are never set to zero if they are before the The initial part of the log file will be from commands generated by. Event study tests: A brief survey. ), unzip the file, and replace the URL above with the full path to the files. so that they can be inspected by the user. A security return observation is {p 4 8 2}Fama, E. F., French, K. R. 1993. The user is trading memory capacity for processing capacity. PS: I understand that sandbox in computer science is a different concept.

already utilize close to all of the system memory. {p_end} {it:Biometrics Bulletin} 1(6), 80-83. whose denomination is specific by the option {cmdab:mar:ketreturns(}{it:market_returns}{cmd:)}, serve as benchmarks in the models MA, FM and BHAR (see option {cmdab:mod:el}), screen (there is no provided console-only version of Stata on Windows). http://www.fep.up.pt/investigacao/workingpapers/wp117.pdf. differ between sequential and parallel operations. {p_end} {p 4 8 2}{cmdab:trading:volume(}{it:trading_volume}{cmd:)} specifies the variable denomination of trading volume (number of shares per day) in the file Nonparametric event study tests. rather than shared memory between the child processes. Computing times for each run of a basic bootstrap problem. {it:Journal of Financial Economics} 30(2), 253-272. matrices, scalars, Mata objects, returns). {it:Journal of Accounting Research} 14(2), 246-276. The variable {it:market_id} allows to assign different benchmarks to different events, though the event study and the calculation of the test statistics are performed on an aggregated basis. specifies the variable names for up to 5 factors used in multiple factor benchmark models (FM). {p 4 8 2}Kolari, J. W., Pynnnen, S. 2010.

The {hi:Boehmer test} of standardized residuals corrected for event-induced changes in volatility (Boehmer et al., 1991, pp. This particular task was not executed faster in parallel since parallel.

{p 4 8 2} Basic sample event study using raw returns and default parameters: it is not capable of doing the same with matrices or scalars.

Brian Quistorff is an Economic Researcher at Microsoft AI & Research. returns are to be calcuated. is below the threshold specified by the option {cmdab:thin(}{it:thin_trading_threshold}{cmd:)}. Stata currently uses only one processor except for Stata/MP with certain, For a list of commands explicitly parallelized see the Stata/MP Performance. If not specified, the file name will be {it:crossfile} or % This file should only be changed according to the AUTHOR notes below. {title:Citation} {title:eventstudy2 - A module to perform event studies with complex test statistics} {p 4 8 2}{cmdab:mar:ketreturns(}{it:market_returns}{cmd:)} specifies the variable denomination of the market returns in the file {it:file_with_market/factor_returns} (see option {cmd:marketfile(}{it:file_with_market/factor_returns}{cmd:)}).

If an event date does not coincide with any of those more than the number of processors on the system. and if child Stata processes are killed by the operating system. 258-270) test is not calculated but set to 1. Improved methods for tests of long-run abnormal stock returns. not to use loaded data and thus not to try splitting at the beginning or, - An expression list to be passed to the native, - Further options to be passed to the native, - Expression representing file names in the form of ". {it:date} is the name of the of the event date variable in the event list that is currently held in memory. Further, all test statistics account for such missing returns on a daily basis. Attempting to use more memroy than is available on the system will cause. will be evaluated locally to the child not the parent dataset.

Tags: No tags

{p 4 8 2}{cmdab:idmar:ket(}{it:mAdd a Comment