Most Recent and Search - Wealth Lab Forum

Cone

#1

6/22/2008 3:16 AM

Although the forum path is displayed, it might not be immediately obvious that the Most Recent and Search results are color-coded by forum. If the result has a blue square in the left-most column, it's comes from the Wealth-Lab .NET forum. Likewise, green indicates the Wealth-Lab 4 forum.

fundtimer

#2

6/22/2008 11:49 AM

I'm noticing that your page structure (url) is set-up to support temporary pages. if I bookmark a page, will that page url remain permanent as with the old site or is it subject to change (I suspect that it wont and thus the forthcoming "My Startpage").

fundtimer

#3

6/22/2008 11:54 AM

A few more (premature) comments:

It would be nice if the Last Post at the Most Recent link was time/date ordered. It's almost there.

The last post user name is not always the last post.

fundtimer

#4

6/22/2008 12:04 PM

I've discovered what might be considered a serious accuracy error in the WL5 (and WL4) StdError of estimate indicator( i.e., std error of the linear reg line). For ver 5, would you suggest adding the corrected indicator code to the Wiki open source database or is there a better method (discussion or whatever), when internal indicator errors are found.

No time to submit anything formal so I'll just dump my findings here. if WLab internals want to investigate further then great, otherwise so be it.

The built-in StdErrorSeries (std of estimate) is off by approx 1 std of the lrs line and does not agree with any of the chart servers or other trading systems supporting this indicator. However, my derivation of the indicator (a Pearson's Population correlation derivation) agrees with all of the latter:

CODE:
Please log in to see this code.

See the following for the formula that I used:
http://davidmlane.com/hyperstat/A120567.html

Enjoy!

Size:

Color:

I've discovered what might be considered a serious accuracy error in the WL5 (and WL4) StdError of estimate indicator( i.e., std error of the linear reg line). For ver 5, would you suggest adding the corrected indicator code to the Wiki open source database or is there a  better method (discussion or whatever), when internal indicator errors are found.

No time to submit anything formal so I'll just dump my findings here. if WLab internals want to investigate further then great, otherwise so be it.

The built-in StdErrorSeries (std of estimate) is off by approx 1 std of the lrs line and does not agree with any of the chart servers or other trading systems supporting this indicator. However, my derivation of the indicator (a Pearson's Population correlation derivation) agrees with all of the latter:[code]function OKStdErrorSeries( p_series, p_period: integer ): integer;
begin
  var sname: string = 'OKStdError('+GetDescription(p_series)+','+ IntToStr(p_period)+')';
  Result := FindNamedSeries( sname );
  if ( Result >= 0 ) then Exit;
  var Bar, X, C, j: integer;
  
  //Population STD:
  var yprime: integer = StdDevSeries( p_series, p_period );
  
  Result := CreateSeries();
  C := CreateSeriesLength( p_period );
  X := CreateSeriesLength( p_Period );
  for Bar := p_period to BarCount-1 do
  begin
    for j := p_period-1 downto 0 do
    begin
      @C[j] := @p_series[Bar-j];
      @X[j] := p_period-1-j;
    end;
    @Result[Bar] := @yprime[Bar]*sqrt( 1-sqr( Correlation( X, C, 0, p_period-1 ) ) );
  end;
  SetDescription( Result, sname );
end;

var p1: integer = CreatePane( 100, -1, -1 );
PlotSeriesLabel( StdErrorSeries( #Close, 5 ), p1, 900, 0, 'wl-stderr-5' );
PlotSeriesLabel( OKStdErrorSeries( #Close, 5 ), p1, 9, 0, 'ok-stderr-5' );

var p2: integer = CreatePane( 100, -1, -1 );
PlotSeriesLabel( StdErrorSeries( #Close, 20 ), p2, 900, 0, 'wl-stderr-20' );
PlotSeriesLabel( OKStdErrorSeries( #Close, 20 ), p2, 9, 0, 'ok-stderr-20' );[/code]

See the following for the formula that I used:
[LINK]http://davidmlane.com/hyperstat/A120567.html[/LINK]

Enjoy!

Cone

#5

6/22/2008 4:47 PM

I think we should get a Support Ticket to follow up on an error with an indicator.

vk

#6

6/26/2008 9:46 AM

Hi fundtimer, sorry for taking so long to comment on your suggestions. Bookmarking is not supported now but will be in near future. The url will remain permanent as with the old site, yes.
Most recent should be as it was on the "old" web site, except that if someone edits his post it becomes the "most recent". We feel that this is more accurate, please let me know if there are reasonable objections?

In regards to:"The last post user name is not always the last post" - we have to investigate that and fix it if that is the case.

fundtimer

#7

6/26/2008 12:35 PM

"Most recent should be as it was on the "old" web site..."

Not exactly. On the old site, the latest most recent was at the top of the list and the earliest most recent was at the bottom. That's 'almost' the case here.

p.s. That appears to have been corrected.

Eugene

#8

7/1/2008 12:46 PM

For ver 5, would you suggest adding the corrected indicator code to the Wiki open source database or is there a better method (discussion or whatever), when internal indicator errors are found.

Although we had plans to eventually lock permanent content like standard indicator documentation to protect from spam or wiki vandals, one can always use the "Discuss" button on a wiki page to share valuable findings like the one that you posted. Thanks anyway.

deepbuyer

#9

7/6/2008 2:35 PM

could not find a way to bookmark a thread so that it is stored in my profile

Eugene

#10

7/6/2008 3:26 PM

"Coming Soon!"

Marcel

#11

8/18/2008 11:34 AM

QUOTE:
could not find a way to bookmark a thread so that it is stored in my profile

Hi deepbuyer,
we are working on this feature at the moment.
We will notify you when this feature is available.

EDIT:
The first version of the bookmark feature is available now.

deepbuyer

#12

8/30/2008 5:07 PM

not sure if this has been reported. after i use search and click on one of the results, i can't come back to the search results (expired) and i need to search again. i am using firefox 3.0.1