topic How to conduct calculation with conditions for target dataset rows? in SAS/IML Software and Matrix Computations

How to conduct calculation with conditions for target dataset rows?

Sasulee — Wed, 17 Jun 2020 12:39:02 GMT

Hi all,

A basic question here. I have a VIX 2014 dataset, which is shown as:

date	tau	K	call_option_price	S	r	put_option_price
2014/1/2	16	1735	98.6	1831.98	0.0018	2.175
2014/1/2	16	1740	93.9	1831.98	0.0018	2.175
2014/1/2	16	1745	88.85	1831.98	0.0018	2.525
2014/1/2	16	1750	84.05	1831.98	0.0018	2.6
2014/1/2	16	1755	79.4	1831.98	0.0018	2.825

...

2014/1/2	36	1740	97.7	1831.98	0.0022	7.6
2014/1/2	36	1745	93.4	1831.98	0.0022	8.05
2014/1/2	36	1750	89	1831.98	0.0022	8.55
2014/1/2	36	1755	84.6	1831.98	0.0022	9.15
2014/1/2	36	1760	80.25	1831.98	0.0022	9.8
2014/1/2	36	1765	75.95	1831.98	0.0022	10.45
2014/1/2	36	1770	71.5	1831.98	0.0022	11.25

...

2014/1/3	15	1735	97.25	1831.37	0.0018	1.625
2014/1/3	15	1740	92.6	1831.37	0.0018	1.775
2014/1/3	15	1745	87.5	1831.37	0.0018	1.925
2014/1/3	15	1750	82.65	1831.37	0.0018	1.925
2014/1/3	15	1755	77.8	1831.37	0.0018	2.25
2014/1/3	15	1760	73	1831.37	0.0018	2.45
2014/1/3	15	1765	68.45	1831.37	0.0018	2.75
2014/1/3	15	1770	63.75	1831.37	0.0018	3.05
2014/1/3	15	1775	59.05	1831.37	0.0018	3.375
2014/1/3	15	1780	54.25	1831.37	0.0018	3.7

...

2014/1/3	35	1740	96.45	1831.37	0.0022	6.9
2014/1/3	35	1745	92.45	1831.37	0.0022	7.3
2014/1/3	35	1750	87.95	1831.37	0.0022	7.85
2014/1/3	35	1755	83.55	1831.37	0.0022	8.35
2014/1/3	35	1760	79.2	1831.37	0.0022	9
2014/1/3	35	1765	74.9	1831.37	0.0022	9.7
2014/1/3	35	1770	70.65	1831.37	0.0022	10.4
2014/1/3	35	1775	66.5	1831.37	0.0022	11.2
2014/1/3	35	1780	62.4	1831.37	0.0022	12.1
2014/1/3	35	1785	58.35	1831.37	0.0022	13.05

The data is imported by proc import. Just focus the variable 'tau', where is defined as the near-term expiration (T1) by tau=16 for 2014/1/2 and next-term expiration (T2) by tau=36 for 2014/1/2. Similarly, the T1 for 2014/1/3 is 15 and T2 for the same day is 35.

The question is, how can I extract the T1 and T2 with corresponding other variables in the same row separately into two different tables or matrixes? and make it is available to conduct matrix calculation, such as i = 0 to n,

Y = a*T1[ i ] * b*T2[ i ] (Just a simple example, I have to conduct a complex calculation then)?

Thank you.

Re: How to conduct calculation with conditions for target dataset rows?

PaigeMiller — Wed, 17 Jun 2020 12:45:28 GMT

Here's a simple example of IML code that subsets the input data set base upon a WHERE clause, that you can modify. It refers to a SAS dataset named SASHELP.CLASS, which you have. https://documentation.sas.com/?cdcId=pgmsascdc&cdcVersion=9.4_3.4&docsetId=imlug&docsetTarget=imlug_langref_sect543.htm&locale=en

use Sashelp.Class var{name sex age} where(age>10);

Alternatively, you can place a WHERE clause in the IML READ statement. https://documentation.sas.com/?cdcId=pgmsascdc&cdcVersion=9.4_3.4&docsetId=imlug&docsetTarget=imlug_langref_sect406.htm&locale=en

read all var {Weight} where(Sex='M'); /* vector of male weights */

Re: How to conduct calculation with conditions for target dataset rows?

Rick_SAS — Wed, 17 Jun 2020 12:50:50 GMT

I think you are asking how to extract the unique values of tau. Basically, this is a form of BY-group analysis. The easiest way (conceptually) is to use the UNIQUE-LOC technique to extract the rows that correspond to each level of the tau variable. The link includes an example.

For your data, you would locate the observations for each level of tau and then extract the relevant rows into a matrix. You didn't supply your IML program, but it might look something like this:

do i = 1 to ncol(u);     /* 4. For each level... */
   idx = loc(tau=u[i]);  /* 5. Find observations in level */
   A = x[idx, ];         /* 6. Extract rows of the X matrix into A */
   /* now process and compute with the A matrix */
end;

It is not clear to me if you want to also read/process each date separately. If so, there is a slightly more complicated version of the UNIQUE-LOC technique that handles multiple BY-group variables.