turn on suggestions

Auto-suggest helps you quickly narrow down your search results by suggesting possible matches as you type.

Showing results for

Find a Community

Topic Options

- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

12-23-2014 12:59 PM

I am striving to calculate Value at risk in financial market, in order to produce 10% VaR in excel it is easy to use Norm.inv(probability,mean,std). what is the same code in IML?

Furthermore in excel we have " small(array,3) " to calculate 3rd smallest value. how is it done in IML?

Please help me in this topic. thanks in advance..

Accepted Solutions

Solution

12-24-2014
07:54 AM

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

12-24-2014 07:54 AM

The inverse CDF function is called the QUANTILE function in SAS. To compute the inverse CDF of the normal distribution, use the following:

proc iml;

prob=0.5; mean=0; std=1;

q = quantile("Normal", prob, mean, std);

For computing the kth smallest, I'd use the RANK function., which returns the order of each element in a sorted list. You can use the LOC function to find the index that contains the kth smallest, and then extract the value, as follows:

x = 100:91;

r = rank(x);

idx = loc(r=3); /* location of the 3rd smallest */

small3 = x[idx]; /* value of the 3rd smallest */

All Replies

Solution

12-24-2014
07:54 AM

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

12-24-2014 07:54 AM

The inverse CDF function is called the QUANTILE function in SAS. To compute the inverse CDF of the normal distribution, use the following:

proc iml;

prob=0.5; mean=0; std=1;

q = quantile("Normal", prob, mean, std);

For computing the kth smallest, I'd use the RANK function., which returns the order of each element in a sorted list. You can use the LOC function to find the index that contains the kth smallest, and then extract the value, as follows:

x = 100:91;

r = rank(x);

idx = loc(r=3); /* location of the 3rd smallest */

small3 = x[idx]; /* value of the 3rd smallest */

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

12-24-2014 08:11 AM

Rick,

What if there are some ties ? How to get rid of them ?

proc iml; m = { 1 2 0, 2 4 0, 10 11 12, 2 2 2}; r=rank(m); x=loc(r=3); smallest3=m; print smallest3; quit;

Xia Keshan

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

12-26-2014 07:38 AM

Why would I want to get rid of them? The algorithm returns the observations with the k_th smallest values. If you just want one, you can use

smallest3=m

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

12-26-2014 08:48 AM

As you can see the code return 1,but the smallest 3 is 2 . so do I get that 2 ?

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

12-26-2014 01:10 PM

Sorry, I didn't understand your question previously. In your example you have two 0s. Your code returns 1 because that is the third value in a ranked list of the values: 0, 0, 1, 2, 2, 2,....

If you want to return all values that have the third largest UNIQUE value (which is 2), then you can use the UNIQUE-LOC technique: The UNIQUE-LOC trick: A real treat! - The DO Loop

The code would look like this:

m = { 1 2 0,

2 4 0,

10 11 12,

2 2 2};

u = unique(m); /* unique values */

k = 3; /* look for the k_th smallest */

idx = loc(m=u

print (idx`)[L="Obs Num"] (m[idx])[L="Value"];

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

12-27-2014 03:18 AM

Well done. Thanks Rick .

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

12-24-2014 09:01 AM

Real Thanks