Date:         Tue, 20 Jun 2006 09:26:21 -0400
Reply-To:     "Howard Schreier <hs AT dc-sug DOT org>" <nospam@HOWLES.COM>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         "Howard Schreier <hs AT dc-sug DOT org>" <nospam@HOWLES.COM>
Subject:      Re: WKret

On Tue, 20 Jun 2006 05:11:21 -0400, Thomas <tythong@YAHOO.COM> wrote:

>On Mon, 19 Jun 2006 08:55:08 -0400, Howard Schreier <hs AT dc-sug DOT org>
><nospam@HOWLES.COM> wrote:
>
>>It would be better to have an example of the desired output and a display
>of
>>the code which has been tried, or at least an explanation of whatever
>>approaches have been considered.
>>
>>Nayway, here's a possibility:
>>
>>   proc sql;
>>   create table want as
>>    select com,
>>           intnx('week.5',date,0) format weekdate. as Thursday,
>>           exp(sum(log(ret+1) ) ) - 1 as compound_ret
>>     from have
>>     group by com, thursday
>>     having weekday(min(date) )=5
>>        and weekday(max(date) )=4;
>>   quit;
>>
>>Result:
>>
>>   Obs    COM              Thursday                ret
>>
>>    1     KO         Thursday, March 4, 1999      0.02028
>>    2     KO        Thursday, March 11, 1999      0.06519
>>    3     KO        Thursday, March 18, 1999     -0.02152
>>    4     MSFT     Thursday, January 7, 1999     -0.04917
>>    5     MSFT    Thursday, January 14, 1999      0.13081
>>    6     MSFT    Thursday, January 21, 1999      0.03689
>>
>>On Sun, 18 Jun 2006 11:56:13 -0400, Thomas <tythong@YAHOO.COM> wrote:
>>
>>>Hi all,
>>>
>>>May I know how to compound the returns (ret) from Thursday to Wednesday
>>>for the following dataset? The daily returns are actually compounded from
>>>Thus to Wed into weekly returns for each company (COM). Your kind help
>>>will be greatly appreciated.
>>>
>>>Thank you very much!
>>>
>>>Best regards,
>>>Thomas
>>>
>>>/*DATASET*/
>>>date COM RET          WEEKDAY
>>>19990104 MSFT 0.016674178 2
>>>19990105 MSFT 0.039007094 3
>>>19990106 MSFT 0.032423209 4
>>>19990107 MSFT -0.004958678 5
>>>19990108 MSFT -0.004152824 6
>>>19990111 MSFT -0.015846539 2
>>>19990112 MSFT -0.036016949 3
>>>19990113 MSFT 0.011428571 4
>>>19990114 MSFT -0.014341591 5
>>>19990115 MSFT 0.056437388 6
>>>19990119 MSFT 0.039232053 3
>>>19990120 MSFT 0.044979919 4
>>>19990121 MSFT -0.026518064 5
>>>19990122 MSFT -0.01302803 6
>>>19990125 MSFT 0.035999998 2
>>>19990126 MSFT 0.059845559 3
>>>19990127 MSFT -0.01712204 4
>>>19990128 MSFT 0.031875465 5
>>>19990129 MSFT 0.005747126 6
>>>19990301 KO -0.002935421 2
>>>19990302 KO -0.024533857 3
>>>19990303 KO -0.008048289 4
>>>19990304 KO 0.021298174 5
>>>19990305 KO 0.015888778 6
>>>19990308 KO 0.000977517 2
>>>19990309 KO -0.015625 3
>>>19990310 KO -0.001984127 4
>>>19990311 KO 0.058210734 5
>>>19990312 KO 0.046139359 6
>>>19990315 KO -0.01080108 2
>>>19990316 KO 0          3
>>>19990317 KO -0.027297543 4
>>>19990318 KO 0.021515435 5
>>>19990319 KO -0.010989011 6
>>>19990322 KO 0.002777778 2
>>>19990323 KO -0.02677747 3
>>>19990324 KO -0.007590133 4
>>>19990325 KO 0.007648183 5
>>>19990326 KO -0.008538899 6
>
>Hi Howard,
>
>Thanks lot! May I know if Thursday is a holiday (i.e, missing observation
>for Thursday), does PROC SQL take the subsequent non-holiday observations
>(i.e., maybe start from Friday, Monday, so on and so for) for compounding
>in order to complete the weekly returns?

SQL of course does what one tells it to do.

I coded to require presence of the end-point days (Thursday and Wednesday)
because I saw in the sample data that there were incomplete
Thursday-to-Wednesday weeks at the beginning and end of each series, and I
assumed that those weeks should be excluded.

You can of course tweak the code as needed; consult the PROC SQL
documentation as necessary.

>I would very much appreciate it, if you could kindly advise me again at
>your convenience.
>
>Thanks again & warm regards,
>Thomas