MaplePrimes - answers and comments on Question, Ploting a 3D graph and make a 3D regresssion

Regression

Robert Israel — Thu, 21 Jan 2010 22:36:03 Z

For annotations on the axes,

> pointplot3d(..., labels=['REDOX,'IC50','LIPO']);

For the regression, assuming you want the best least-squares fit for a plane

LIPO = a*REDOX + b*IC50 + c

you can use LSSolve in the Optimization package:

> Optimization[LSSolve]([seq(LIPO[i]-(a*REDOX[i]+b*IC50[i]+c), 
        i=1..7)]);

> plane:= subs(%[2],a*x + b*y + c);

NonlinearFit

herclau — Fri, 22 Jan 2010 01:52:26 Z

I hope this helps ...

f := proc (p, r, i) r[i, 3]-p[1]*r[i, 1]-p[2]*r[i, 2]-p[3] end proc;
H := Statistics[NonlinearFit](3, f, <<(REDOX)>|<IC50>|<LIPO>>, 
    output = solutionmodule);

H:-Results();

 ["residualmeansquare" = 0.0626631103666510,

 "residualsumofsquares" = 0.250652441466604,

 "residualstandarddeviation" = 0.250326008170647,

  "degreesoffreedom" = 4,

   "parametervalues" = ...

   "parametervector" = ...

   "residuals" = ...

You could see a bit more with this:


infolevel[Optimization] := 10;
 infolevel[Statistics] := 3;

I thank you very much for

Bibinou — Sat, 23 Jan 2010 15:47:06 Z

I thank you very much for your prompt answer,

but I don't really understand all mathematical concept there is behind... Indeed I'm a French chemist and I'm not very used to use Maple software.

I tried to modify the code to represent my 3 experimental variables

>with(plots);
> IC50 := [5, .5, .6, 1.5, 1, 3, .8];
> REDOX := [.2, .41, .52, .72, .69, .56, .38];
> LIPO := [.572, 2.233, 1.966, 2.107, 1.919, 1.008, 2.345];
> s := NULL;
> for i to nops(IC50) do s := s, [LIPO[i], REDOX[i], IC50[i]] end do;
> pts := [s];
> pointplot3d(pts, axes = normal, axis[2] = [color = grey], axis[3] = [color = red], thickness = 3, font = [HELVETICA, BOLD, 11], title = "Correlation IC50 REDOX LIPOPHILIE", color = blue, symbol = sphere);

but I think there's no difference in the representation. Isn't it?

1) I tried to add label in pointplot3d function, but Maple returns an error.

> pointplot3d(pts,axes=normal, axis[2]=[color=grey], axis[3]=[color=red],thickness=3,font=[HELVETICA,BOLD,11],title="Correlation IC50 REDOX LIPOPHILIE", color=blue,symbol=sphere, labels=['REDOX,'IC50','LIPO']);

Error, unable to match delimiters

2) I copy/paste the procedure from Herclau (without understanding so much....) but this don't give me a graph...? isn't it?

I read the help file for the function CurveFitting[ArrayInterpolation] and I think this could help me to draw my graph. But I really don't understand how to use it with my variable given that each variable is independant from another. The result of my 3D map should be a kind of sink.

Thank you for your help.

For 1), try

gkokovidis — Sun, 24 Jan 2010 06:19:34 Z

For 1), try this:

pointplot3d(pts,axes=normal, axis[2]=[color=grey], axis[3]=[color=red],thickness=3,font=[HELVETICA,BOLD,11],title="Correlation IC50 REDOX LIPOPHILIE", color=blue,symbol=sphere, labels=["REDOX","IC50","LIPO"]);

You must be using the Standard interface, because with the Classic interface, the symbol=sphere option is not recognized. The code above works in Standard with v12.

Regards,
Georgios Kokovidis

Dräger Medical

a gpaph

herclau — Tue, 26 Jan 2010 19:11:10 Z

this give me a graph...

Statistics[Fit](p1*x1+p2*x2+p3, independentDataMatrix, LIPO, [x1, x2]);

-0.399224485713786436 x1 - 0.239079452680029364 x2 + 2.56176858831106546

g1 := plots[pointplot3d](REDOX, IC50, LIPO);

plots[display](plot3d(-.399224485713786436*x1-.239079452680029364*x2+2.56176858831106546, x1 = -1 .. 1, x2 = 0 .. 5, color = grey, style = wireframe), g1, axes = normal, axis[2] = [color = grey], axis[3] = [color = red], thickness = 3, font = [HELVETICA, BOLD, 11], title = "Correlation IC50 REDOX LIPOPHILIE", color = blue, symbol = solidsphere, labels = ["REDOX", "IC50", "LIPO"]);

Indeed it works, but that's

Bibinou — Thu, 28 Jan 2010 14:19:56 Z

Indeed it works, but that's not exactly what I expected. I would like not to have a simple line, but a kind of surface delimitated by my different points. Thanks a lot for your help