Skip to content

quantstrat spread parameter sweep problem

3 messages · Brian G. Peterson, Rob Schmidt

Original
Rob Schmidt
# 
Hello all,

I am stuck on making a parameter sweep of a spread and have attached 
the version that gets the farthest before failing.  This has driven me 
to extreme measures like reading the documentation and trying to upgrade 
my debugging skills.  

The code gets through the basic calculations and makes a chart so the 
spread definitions seem ok.  But it fails in applyParameter.  The variable 
values are reasonable as far as I can tell until %dopar% in the foreach() 
command in parameters.R.  Then the debugger takes the rest of the 
command in {} as a big chunk and my modest debug skills fail.  This 
sweep code works when I use a single contract rather than a spread.  
Here is the error.  The traceback points to the chunk of code after 
%dopar% so I won't repeat that.

Thanks to the authors for this great effort. Thanks for all help.

Best regards,

Rob
[1] "2011-03-01 00:00:00 SPY.DIA -100 @ 10.9387968680294"
[1] "2011-03-07 00:00:00 SPY.DIA 100 @ 10.8600842370335"
[1] "2011-03-10 00:00:00 SPY.DIA -100 @ 9.82580786782957"
[1] "2011-03-11 00:00:00 SPY.DIA 100 @ 10.0274530275673"
[1] "2011-03-15 00:00:00 SPY.DIA -100 @ 9.2739827055253"
[1] "2011-04-06 00:00:00 SPY.DIA 100 @ 9.78335984354628"
[1] "2011-04-13 00:00:00 SPY.DIA -100 @ 9.02241069100391"
Realized PL: -63.23097 
[1] "Object for parameter distribution initialized..."
[1] "Parameter constraint object initialized..."
[1] "ParamTable generated"
numValues: 5, numResults: 0, stopped: TRUE
got results for task 1
accumulate got an error result
numValues: 5, numResults: 1, stopped: TRUE
returning status FALSE
got results for task 2
numValues: 5, numResults: 2, stopped: TRUE
returning status FALSE
got results for task 3
numValues: 5, numResults: 3, stopped: TRUE
returning status FALSE
got results for task 4
numValues: 5, numResults: 4, stopped: TRUE
returning status FALSE
got results for task 5
numValues: 5, numResults: 5, stopped: TRUE
not calling combine function due to errors
returning status TRUE
Error in { (from spread-sweep05.R#193) : task 1 failed - "argument is of
length zero"
R version 2.15.3 (2013-03-01)
Platform: x86_64-pc-linux-gnu (64-bit)

locale:
 [1] LC_CTYPE=en_US.UTF-8       LC_NUMERIC=C              
 [3] LC_TIME=en_US.UTF-8        LC_COLLATE=en_US.UTF-8    
 [5] LC_MONETARY=en_US.UTF-8    LC_MESSAGES=en_US.UTF-8   
 [7] LC_PAPER=C                 LC_NAME=C                 
 [9] LC_ADDRESS=C               LC_TELEPHONE=C            
[11] LC_MEASUREMENT=en_US.UTF-8 LC_IDENTIFICATION=C       

attached base packages:
[1] parallel  stats     graphics  grDevices utils     datasets  methods  
[8] base     

other attached packages:
 [1] doMC_1.3.0                 iterators_1.0.6           
 [3] quantstrat_0.7.7           foreach_1.4.0             
 [5] blotter_0.8.14             PerformanceAnalytics_1.1.0
 [7] FinancialInstrument_1.1.7  quantmod_0.4-0            
 [9] Defaults_1.1-1             TTR_0.22-0                
[11] xts_0.9-3                  zoo_1.7-9                 
[13] vimcom_0.9-8               setwidth_1.0-3            

loaded via a namespace (and not attached):
[1] codetools_0.2-8 compiler_2.15.3 grid_2.15.3     lattice_0.20-13
[5] tools_2.15.3

spread-sweep05.R
<http://r.789695.n4.nabble.com/file/n4664819/spread-sweep05.R>  



--
View this message in context: http://r.789695.n4.nabble.com/quantstrat-spread-parameter-sweep-problem-tp4664819.html
Sent from the Rmetrics mailing list archive at Nabble.com.
Brian G. Peterson
#
On 04/20/2013 09:29 AM, Rob Schmidt wrote:
%dopar% is running in parallel.  Obviously you can't get a debugger down 
into the child processes (debugging for parallel computing is an even 
darker art than R coding).

You may be able to debug into your problem by calling registerDoSEQ() 
before running your script, to force foreach into single threaded mode 
so that you can debug into the foreach loop.
This is saying the error is on line 193 of your script, but I suspect 
that you're going to come back and say that is the applyParameters line.

Now that I've said all that, we've moved all our development attention 
on parameter code to apply.paramsets instead of applyParameters.  The 
old code wasn't extensible enough, and we've basically abandoned it in 
favor of a more flexible apporoach, and a refactored code base.

You can see all the documentation, and you should be able to find the 
code in the paramsets.R file in the source.  One of the luxor demos uses 
the paramsets code, so you shjould be able to find example code to start 
from as well in modifying your parameter specification.

Regards,

Brian
1 day later
Rob Schmidt
# 
Hi Brian,

Thank you for your helpful suggestions.  The paramset method is working much
better.  For completeness, I'm attaching a minimum parameter sweep example
using apply.paramset() based on a simple moving average crossover.  It is a
mangle of a small piece of the luxor demo.  It does make a warning about
GBPUSD being the name of a currency about which I have no clue.

Best regards,

Rob

spread-sweep08.R
<http://r.789695.n4.nabble.com/file/n4664943/spread-sweep08.R>  



--
View this message in context: http://r.789695.n4.nabble.com/quantstrat-spread-parameter-sweep-problem-tp4664819p4664943.html
Sent from the Rmetrics mailing list archive at Nabble.com.