Hi,

Forget about the Linear Regression.  I can use Rsquared: I'd like to be able
to calculate for each bar:

a) if it is close to a very good rsquared fit (let's say +0.50);
b) if the line that is drawn (or calculated, since there is no need to draw
it) has a slope between 10 and 20 degrees.

How would you do that?

Thanks,

Louis


2008/9/13 gp_sydney <[EMAIL PROTECTED]>

>   Sorry, I'm not up on the maths of regression, and am not sure what
> that code is trying to do. It seems to be correlating the Close array
> to a fixed-slope bar number array (BarIndex) over the range since the
> most recent HHV. From what I can see, there is no use of LastValue or
> SelectedValue in that code, and so as long as MA and Sum can accept
> arrays as the range, then r2 should be an array with the appropriate
> value at each bar.
>
> However, I'm not sure what correlating to the BarIndex array is
> achieving, or what you mean by drawing (or calculating) lines from
> some point to the HHV or by backtesting with a particular gradient.
> Perhaps it's obvious, but regression is not something I'm very
> familiar with.
>
>
> Regards,
> GP
>
> --- In [email protected] <amibroker%40yahoogroups.com>, "Louis P."
> <[EMAIL PROTECTED]> wrote:
> >
> > Hi,
> >
> > Ok let's look at this from a different perspective:
> >
> > // variable-period Correlation
> > function CorrV( x, y, range )
> > {
> >
> > avX = MA( x, range );
> > avY = MA( y, range );
> > avXY = MA( x * y, range );
> >
> > ssqrX = Sum( x * x, range );
> > ssqrY = Sum( y * y, range );
> >
> > VarX = Max( ssqrx/range - avX * avX, 0 );
> > VarY = Max( ssqry/range - avY * avY, 0 );
> >
> > return Nz( ( avXY - avX * avY ) / sqrt( VarX * VarY ) );
> > }
> >
> > // variable-period R squared
> > function RSquaredV( array, range )
> > {
> > return CorrV( BarIndex()+1, array, range ) ^ 2;
> > }
> >
> > Periods = Param("HHV Period", 20, 1, 100, 1 );
> >
> > // bars since HHV
> > barhh = HHVBars( High, Periods );
> >
> > r2 = RSquaredV( C, Max( 1, barhh ) );
> >
> > With the code above, I get a Rsquared from a HHV, which is good.
> But I'd
> > like that to be backtestable only with a slope of a particular gradient,
> > let's say 10-20 degrees. I know it would sound useless to be able
> to draw a
> > line from each point to the HHV before (the chart would look like a
> spider
> > web) and I don't want to draw the stuff. But I'd like to be able to
> > calculate for each bar:
> >
> > a) if it is close to a very good rsquared fit (let's say +0.50);
> > b) if the line that is drawn (or calculated, since there is no need
> to draw
> > it) has a slope between 10 and 20 degrees.
> >
> > How would you do that?
> >
> > Thanks,
> >
> > Louis
> >
> >
> >
> > 2008/9/12 gp_sydney <[EMAIL PROTECTED]>
> >
> > > I think the first question needs to be what is the result you are
> > > expecting?
> > >
> > > In that code you are passing arrays to functions that expect single
> > > values, which is why the SelectedValue and LastValue functions would
> > > have been there before.
> > >
> > > LineArray draws a line between two points on the chart, and the
> > > coordinates of those points need to be single values. What would you
> > > expect to see by passing arrays as the coordinates?
> > >
> > > Regards,
> > > GP
> > >
> > >
> > > --- In [email protected] <amibroker%40yahoogroups.com><amibroker%
> 40yahoogroups.com>,
>
> "Louis P."
> > > <rockprog80@> wrote:
> > > >
> > > > Hi,
> > > >
> > > > Here is what I tried to do:
> > > >
> > > > barhh1 = HHVBars( High, Periods ) ;
> > > > bi1 = BarIndex() ;
> > > > y11 = LinearReg( C, barhh1 ) ;
> > > > y01 = LinRegIntercept( H, barhh1 ) ;
> > > > sl1 = LineArray( bi1-barhh1+0, y01, bi1, y11, 0, True );
> > > >
> > > > Unfortunately it seems there is an error...
> > > >
> > > > What's the barindex for in that code, and if the problem is there,
> > > how can I
> > > > deal with it, considering that for the first 4 lines of that code it
> > > began
> > > > with "selectedvalue"?
> > > >
> > > > Thanks,
> > > >
> > > > Louis
> > > >
> > > >
> > > >
> > > > 2008/9/12 Barry Scarborough <razzbarry@>
> > > >
> > > > > Ara said just use the variable. For instance C is an array and
> using
> > > > > C will point to the last value, or the current bar, in the
> array. To
> > > > > point to a bar prior to the current one use Ref(C, -n); where n is
> > > > > the number of bars prior to the current one. Replace C with
> the array
> > > > > you want to get the value for.
> > > > >
> > > > > Use LastValue when you need a value and not an array. An
> example is
> > > > > in an IF statement. IF() cannot take an array and you have to
> get the
> > > > > value. If you want the value of some previous value use
> LastValue(ref
> > > > > (c,-n));
> > > > >
> > > > > If your results get confusing dump it to the interpretation
> window so
> > > > > you can compare the values with what you expect them to be. Use
> > > > > something like this:
> > > > >
> > > > > printf(numtostr(lastvalue(c), 1.2)); // current value of c
> > > > > printf("\n" + numtostr(lastvalue(ref(c, -1)), 1.2)); // previous c
> > > > >
> > > > > Barry
> > > > >
> > > > >
> > > > > --- In [email protected] <amibroker%40yahoogroups.com>
> <amibroker%40yahoogroups.com><amibroker%
> > > 40yahoogroups.com>, "Ara
> > > > > Kaloustian" <ara1@> wrote:
> > > > > >
> > > > > > just use the variable (without selectedvalue or lastvalue)
> > > > > > ----- Original Message -----
> > > > > > From: Louis P.
> > > > > > To: [email protected] <amibroker%40yahoogroups.com>
> <amibroker%40yahoogroups.com><amibroker%
> > > 40yahoogroups.com>
> > > > > > Sent: Friday, September 12, 2008 2:47 PM
> > > > > > Subject: [amibroker] Converting selectedbar and lastvalue to...
> > > > > >
> > > > > >
> > > > > > Hi,
> > > > > >
> > > > > > I need to know how to convert code that is using selectedbar (it
> > > > > gets the bar before where I am on the chart) and lastvalue (which
> > > > > gets the bar at the end of the array) to a code that is
> checking each
> > > > > and every bar for the pattern.
> > > > > >
> > > > > > Anyone know how to do that?
> > > > > >
> > > > > > Thanks,
> > > > > >
> > > > > > Louis
> > > > > >
> > > > >
> > > > >
> > > > >
> > > >
> > >
> > >
> > >
> >
>
>  
>

Reply via email to