4
$\begingroup$

I have this system of three equation that mathematica does not resolve.

x = 1005790/1000000;
y = 3012034/1000000;
varx = 9979050/1000000;
vary = 12985206/1000000;
covxy = 9986968/1000000;
z = 3614929/1000000;
varz = 25466937/1000000;
z3 = 323087800/1000000;
eq1 = z == av + at*x + a1*y 
eq2 = varz == at^2 *varx + a1^2 *vary + 2*covxy *a1*at 
eq3 = (av + x*at + a1*y)^3 +3*(av + x*at + a1*y)*(at^2 *varx + a1^2 * vary + 2* at*a1* covxy) == z3 
FindRoot[{eq1 , eq2, eq3}, {{av, 0.4}, {at, 0.7}, {a1, 1.1}}]
Reduce[{eq1 , eq2, eq3}, {av, at, a1}]

Reduce return False and FindRoot is printing the error: FindRoot::jsing: Encountered a singular Jacobian at the point {av,at,a1} = {0.4,0.7,1.1}. Try perturbing the initial point(s). But even if I change the initial points by a Little, The program is unable to solve the equations. I have plotted the three equations minus the known term, and graphically it seems that there is a solution (I fixed av to the correct value, and I see that the three functions are near zero at the same point).

code to plot:

prim[av_, at_, a1_] := av + at*x + a1*y - z; 
sec [av_, at_, a1_] := at^2 *varx + a1^2 *vary + 2*covxy *a1*at - varz;
terz [av_, at_, a1_] := (av + x*at + a1*y)^3 + 
   3*(av + x*at + a1*y)*(at^2 *varx + a1^2 * vary + 
      2* at*a1* covxy) - z3;
Plot3D[{prim[0.1, at, a1], sec[0.1, at, a1], terz[0.1, at, a1]}, {at, 
  0.4, 0.6}, {a1, 0.9, 1.1}]
$\endgroup$
3
  • 4
    $\begingroup$ Can you add the code you used to plot the functions and the value of av you tried? The fact that Reduce returns False seems a strong indication that there is no solution. $\endgroup$
    – MarcoB
    Commented Aug 2, 2022 at 12:47
  • $\begingroup$ Sure! I changed the question $\endgroup$ Commented Aug 2, 2022 at 13:12
  • 2
    $\begingroup$ There is no actual solution (the system is inconsistent). $\endgroup$ Commented Aug 2, 2022 at 19:17

1 Answer 1

4
$\begingroup$

Try NMinimize:

eq = {-z + av + at*x + a1*y,-varz + at^2*varx + a1^2*vary + 2*covxy*a1*at, (av + x*at + a1*y)^3 +3*(av+ x*at + a1*y)*(at^2*varx + a1^2*vary +2*at*a1*covxy) - z3};

sol=NMinimize[eq . eq, {a1, at, av}]
(*{8.30607*10^-6, {a1 -> 0.966855, at ->0.539657, av -> 0.157078}}*)
eq /. sol[[2]] 
(* {-0.00286933, -0.000269104, 0.0000248339}*)

addendum

If we look examplary for positive solutions, two ContourPlot's show

sol1 = Solve[eq[[2]] == 0, a1][[-1]]
Column[{
  ContourPlot[eq[[2]] == 0, {at, 0, 2}, {a1, 0, 2}, 
   FrameLabel -> {at, a1}], 
  ContourPlot[
   Evaluate[Thread[eq[[{1, 3}]] == 0 /. sol1]], {at, 0, 2}, {av, 0, 
    2}, ContourStyle -> {Blue, {Dashed, Red}}, 
   FrameLabel -> {at, av}]}]

enter image description here

that solutions a1[at],av [at]exist for ~.5<at<1.5

$\endgroup$
2
  • $\begingroup$ Thank you! You have found exactly what I needed. The coefficients of the system are statistics, so the correct way is to minimize with the three equation. Now It's clear. Thanks $\endgroup$ Commented Aug 2, 2022 at 13:32
  • $\begingroup$ @AdrianoDelVincio You're welcome. See my addendum! $\endgroup$ Commented Aug 2, 2022 at 13:42

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.