MathGroup Archive 2007

[Date Index] [Thread Index] [Author Index]

Search the Archive

A comparison and thanks

  • To: mathgroup at smc.vnet.net
  • Subject: [mg75464] A comparison and thanks
  • From: "actuary at mchsi.com" <actuary at mchsi.com>
  • Date: Wed, 2 May 2007 03:46:28 -0400 (EDT)

To: All

After sending my initial email, I decided to use approach #3.  After
receiving all of the responses to me question, I ran a comparison of
my method against the methods using the UnitStep function and the
THread function

The problem set-up is that is have two 2-dimensionanl matrices. Each
matrix has 10,000 rows and 359 columns.  I wanted to find, for a given
i the number of times a(i,j) > {b(i,j)+k) and then determine the value
for specific statistic over all 10000 i's.

1.  Timing[Quantile[Table[Total[UnitStep[srl[[i,All]] - (lrl[[i, All]]
+ .15)]], {i, 1, n}], .5]/359 // N]    {2.574 Second, 0.142061}

2.  Timing[Quantile[Table[Count[Thread[srl[[i,ll]] > (lrl[[i, All]] + .
15)], True], {i, 1, n}], .5]/359 // N]  {4.536 Second, 0.1420

3.  Timing[Quantile[Table[Count[srl[[i, All]] - (lrl[[i, All]] + .15),
_?Positive], {i, 1, n}], .5]/359 // N]  61} (5.929 Second, 0.142061}

The UnitStep function did the task the fastest.

Thanks again for all of your suggestions.

Larry



  • Prev by Date: Re: Logical comparisons of items in a two lists
  • Next by Date: averaging sublists of different lengths
  • Previous by thread: Re: MathML 3 working group
  • Next by thread: averaging sublists of different lengths