ValueError: setting an array element with a sequence. when plotting R data

[Rolf]

There is a well known problem (http://ask.sagemath.org/question/192/
compiling-r-with-png-support) with plotting in R that apparently some
people successfully eliminated by installing the libraries mentioned
and recompiling. I didn’t succeed, so I'm trying a work around and
stumbled over an other problem.

If I use this little script:

def ndistribpts(x): return (x,r.dnorm(x,mean=100,sd=25))
v=map(ndistribpts,range(98,103))
v

[(98, [1] 0.01590671), (99, [1] 0.01594493), (100, [1] 0.01595769),
(101, [1] 0.01594493), (102, [1] 0.01590671)]
However this cannot be plotted.

A similar script:

def cube(x): return (x,x*x*x)
v=map(cube,range(98,103))
v

[(98, 941192), (99, 970299), (100, 1000000), (101, 1030301), (102,
1061208)]
That easily can be used by scatter_plot

I suppose it has something to do with the “[1] ” in front of each R
result. I have no idea how to eliminate that. The result cannot be
treated as a

string [3:]

or as matrix .GetItem(1)

Thank you for considering this.
R.

[Rolf]

Problem solved!

The simple '._sage_()' does the trick. It converts data from R into
something that can be handled by Python or Sage.

def ndistribpts(x): return (x,r.dnorm(x,mean=100,sd=25)._sage_())
def cube(x): return (x,x*x*x)
v=map(ndistribpts,range(98,103))
scatter_plot(v)

Works though I'm still far from a pretty curve, but we are improving
rapidly.

Thank you.
R.

[kcrisman]

On Jun 7, 7:08 am, Rolf <kamha...@googlemail.com> wrote:
> Problem solved!
>
> The simple '._sage_()' does the trick. It converts data from R into
> something that can be handled by Python or Sage.

Glad you found a solution!

> def ndistribpts(x): return (x,r.dnorm(x,mean=100,sd=25)._sage_())
> def cube(x): return (x,x*x*x)
> v=map(ndistribpts,range(98,103))
> scatter_plot(v)
>
> Works though I'm still far from a pretty curve, but we are improving
> rapidly.
>

Does this give you what you are looking for? I don't know that you
will get a pretty curve with only a few points, of course, so I
broadened the range.


sage: v=map(ndistribpts,range(50,151))
sage: line(v,ymin=0,ymax=.02)

The command 'points' might also give what you want, without connecting
the points.

- kcrisman

[Rolf]

> sage: v=map(ndistribpts,range(50,151))
> sage: line(v,ymin=0,ymax=.02)
>
> The command 'points' might also give what you want, without connecting
> the points.

Thank you for this enhancement. The curve looks nice. However only in
this example the bell-shape can be appreciated.

It doesn't work here.

def ndistribpts(x): return (x,r.dnorm(x,mean=5,sd=0.5)._sage_())
v=map(ndistribpts,range(1,10))
line(v,ymin=0,ymax=1)

It needs more ajustments than just ymax as the integer steps provided
by the range function are too clumsy.
Thus I wonder why this doesn't work

def f(x): return r.dnorm(x,mean=100,sd=25)._sage_()
plot(f(x)(x,90,110))

while

def f(x): return x^3
plot(f(x),(x,0,2))

works perfectly. Is this a problem that can be solved?

[D. S. McNeil]

> Thus I wonder why this doesn't work
>
> def f(x): return r.dnorm(x,mean=100,sd=25)._sage_()
> plot(f(x)(x,90,110))
>
> while
>
> def f(x): return x^3
> plot(f(x),(x,0,2))
>
> works perfectly.

The problem is that the "f(x)" calls f with the symbolic argument "x"
_at the time you call the plot function_.

In the first case, r.dnorm doesn't know what to do with a symbolic x,
and barfs before it returns. OTOH, in the second function, the f(x)
call simply returns x^3, so it's equivalent to typing plot(x^3, (x, 0,
2).

Both cases should work if you use

plot(f, (x,90,110))
plot(f, (x,0,2))

Doug

[Rolf]

> Both cases should work if you use
>
> plot(f, (x,90,110))
> plot(f, (x,0,2))
>

... and in fact they do.
I feel, this should be documented somewhere.
I used http://tutorial.sagenb.org/pub/4/
for reference.

Best regards
R.

[Jason Grout]

On 6/8/11 5:04 AM, Rolf wrote:
>
>
>> Both cases should work if you use
>>
>> plot(f, (x,90,110))
>> plot(f, (x,0,2))
>>
> ... and in fact they do.
> I feel, this should be documented somewhere.

Is item 4 of http://www.sagemath.org/doc/tutorial/tour_functions.html
sufficient? If not, that might be a place someone could clarify better.

Thanks,

Jason

[kcrisman]

On Jun 8, 6:04 am, Rolf <kamha...@googlemail.com> wrote:
> > Both cases should work if you use
>
> > plot(f, (x,90,110))
> > plot(f, (x,0,2))
>
> ... and in fact they do.
> I feel, this should be documented somewhere.

The difference between symbolic callable functions and all other
function types (not to mention symbolic expressions that are not
functions) is a subtle one. Basically, the rule is that anything you
don't define using the syntax

f(x)=expression_with_x

should be called in plot with

plot(f,(x,90,110))

In particular, anything defined using 'def' will have to be done this
way. Or

plot(f,90,110)

or even

var('y')
plot(f,(y,90,110))