Re: [ferret_users] inconsistent behaviour trying to use min to find location.

[Ansley Manke <ansley.b.manke@xxxxxxxx>]

Hi Rob,
Russ is right that this was a matter of the precision of the numbers. By 
default the grave-accent syntax to evaluate an expression within the 
command line gives you back 4 digits when it evaluates an expression, 
and that wasn't enough in this instance. In the command where you asked for

yes? list FUNDHS50[T=@LOC:27.53]

and get the null result, Ferret is telling you that the expression never 
goes down as low as 27.53, and that's true, based on your listing where 
you used the F9.4 format to view the data with its full precision.  The 
smallest value is 27.5387.   What Russ suggested uses all of the 
precision in the numbers, not making any exact evaluation until it asks 
for the expression minus its min, at zero.  So that is a more robust way 
of doing this.

You could still use the v[x=@loc:`vmax`] syntax, but ask for more 
precision in the output of the grave-accent expression, by saying 
`vmax,prec=6` or 7.

I don't think this behavior would be any different depending on what 
installation of Ferret you are using.

Ansley


On 3/19/2012 5:06 AM, rdbriggs wrote:
> Thanks Russ,
>
> Neat solution. I was stuck in the trees. Strange behaviour though; I 
> have used the other approach numerous times without issue, furthermore 
> I do large O(3k) ensemble analysis of runs use the min function as 
> part of my routine to assess each run. I would have thought I would 
> have hit this issue before. Developers, I can supply the files and my 
> set-up if you have an interest.
>
> regards,
>
> Rob
>
>
>
> On 12-03-18 09:11 PM, Russ Fiedler wrote:
> > Hi Rob,
> >
> > I wonder if there is some floating point rounding going in the immediate
> > mode evaluation that you can't see (i.e. down at the level of precision)
> > and the minimum gets missed.
> >
> > try instead
> >
> > let newvar = var - var[t=@min] ! minima at 0
> > let xloc = var[t=@loc:0]
> >
> > Russ
> >
> >
> > On Mon, 2012-03-19 at 07:49 +1100, rdbriggs wrote:
> > > Hello,
> > >
> > > In the past I've used the
> > >
> > > let vmax = v[x=@max,y=@max]
> > >
> > > and
> > >
> > > let xloc = v[x=@loc:`vmax`]
> > >
> > > commands to successfully find the location of the minimum. But this 
> > > time I have a problem, can anyone tell me what I am doing wrong?
> > >
> > > My functions are FUNDHS, FUNDHS50
> > >
> > > ! get min
> > > LET outDHS = funDHS[T=@MIN]
> > > LET outDHs50 = funDHs50[T=@MIN]
> > >
> > > ! and time it occurs
> > > LET TDHS= funDHS[T=@LOC:`outDHS`]
> > > LET TDHs50 = funDHs50[T=@LOC:`outDHs50`]
> > >
> > > LIST outDHS, outDHS50, TDHs, TDHs50
> > >                DATA SET: ./rreconHs.dat
> > >                T: -120.5 to 0.5
> > >    Column  1: OUTDHS is FUNDHS[T=@MIN]
> > >    Column  2: OUTDHS50 is FUNDHS50[T=@MIN]
> > >    Column  3: TDHS is FUNDHS[T=@LOC:0.7081349]
> > >    Column  4: TDHS50 is FUNDHS50[T=@LOC:27.53873]
> > >             OUTDHS  OUTDHS5  TDHS  TDHS50
> > > I / *:     0.7081   27.54 -10.00   ....
> > >
> > >
> > > ------>   SoTDHS  worksTDHS50  does not ?
> > >
> > > The output from my two functions are
> > >
> > > yes? list FUNDHS, FUNDHS50
> > >
> > > 15    / 106:     50.    163.
> > > -14    / 107:     33.    130.
> > > -13    / 108:     20.    109.
> > > -12    / 109:     10.    104.
> > > -11    / 110:      3.     32.
> > > -10    / 111:      1.     28.<   ------ I want to automatically 
> > > identify this time
> > > -9     / 112:      2.     28.
> > > -8     / 113:      7.     32.
> > > -7     / 114:     16.     41.
> > > -6     / 115:     28.     52.
> > > -5     / 116:     44.     68.
> > > -4     / 117:     64.     87.
> > > -3     / 118:     88.    110.
> > > -2     / 119:    115.    137.
> > > -1     / 120:    145.    167.
> > > 0      / 121:    180.    201.
> > > yes? list FUNDHS, FUNDHS50
> > >
> > >
> > > improving the precision
> > > yes? list/form=(2()) FUNDHS, FUNDHS50
> > >
> > >    33.1926 130.2298
> > >     19.8315 108.5893
> > >     10.4672 104.3460
> > >      2.9604  32.1601
> > >      0.7081  27.5387<------ these are the mins
> > >      2.1548  28.4311
> > >      7.2379  32.3409
> > >     15.9903  40.5858
> > >     28.3927  52.3871
> > >     44.4533  68.0784
> > >     64.1537  86.9202
> > >     87.5170 109.8389
> > >    114.5442 136.8729
> > >    145.2063 166.8670
> > >    179.5263 200.7330
> > >
> > >
> > > yes? list FUNDHS50[T=@LOC:27.53873]
> > >                VARIABLE : T of 27.54 in ((HS_ADJDHS50 - 
> > > HSOBS)/SIGHS50)^2 + ((TAX-TOBS) / TOBS_SIG)^2
> > >                FILENAME : rreconHs.dat
> > >                T        : -120.5 to 0.5 (location of 27.54)
> > >           ....
> > > yes? list FUNDHS50[T=@LOC:28]
> > >                VARIABLE : T of 28 in ((HS_ADJDHS50 - 
> > > HSOBS)/SIGHS50)^2 + ((TAX-TOBS) / TOBS_SIG)^2
> > >                FILENAME : rreconHs.dat
> > >                T        : -120.5 to 0.5 (location of 28)
> > >            -10.10
> > >
> > > yes? list outDHS50
> > >                VARIABLE : FUNDHS50[T=@MIN]
> > >                FILENAME : rreconHs.dat
> > >                T        : -120.5 to 0.5
> > >             27.54
> > > yes? list FUNDHS50[T=@LOC:27.53]
> > >                VARIABLE : T of 27.53 in ((HS_ADJDHS50 - 
> > > HSOBS)/SIGHS50)^2 + ((TAX-TOBS) / TOBS_SIG)^2
> > >                FILENAME : rreconHs.dat
> > >                T        : -120.5 to 0.5 (location of 27.53)
> > >           ....
> > >
> > > Any ideas what is going on here ?
> > >
> > > Regards,
> > >
> > > Rob