[Date Index] [Thread Index] [Author Index]

Re: Solve yields unreal solutions, NSolve yields real solutions

• To: mathgroup at smc.vnet.net
• Subject: [mg58295] Re: Solve yields unreal solutions, NSolve yields real solutions
• From: Jean-Marc Gulliet <jeanmarc.gulliet at gmail.com>
• Date: Sun, 26 Jun 2005 01:33:57 -0400 (EDT)
• Organization: The Open University, Milton Keynes, England
• References: <d9issf$cps$1@smc.vnet.net>
• Sender: owner-wri-mathgroup at wolfram.com

Mukhtar Bekkali wrote:
> I thought NSolve is the same as Solve just numeric. The below code
> gives real answers, replacing NSolve with Solve in the last line
> generates Unreal results.  I do not understand why.
> 
> \!\(\*
>   RowBox[{\(Clear["\<Global`*\>"]\), "\n", \(Remove["\<Global`*\>"]\),
>     "\[IndentingNewLine]", \(Needs["\<Miscellaneous`RealOnly`\>"]\),
>     "\[IndentingNewLine]",
>     RowBox[{
>       RowBox[{
>         RowBox[{
>           StyleBox["{",
>             FontColor->RGBColor[1, 0, 1]],
>           StyleBox[\(\[Lambda], \[Gamma], k\),
>             FontColor->RGBColor[1, 0, 1]],
>           StyleBox["}",
>             FontColor->RGBColor[1, 0, 0]]}],
>         StyleBox["=",
>           FontColor->RGBColor[1, 0, 0]],
>         RowBox[{
>           StyleBox["{",
>             FontColor->RGBColor[1, 0, 0]],
>           RowBox[{"1",
>             StyleBox[",",
>               FontColor->RGBColor[1, 0, 0]], \(2\/3\), ",", "0"}],
> "}"}]}],
>       ";"}], "\[IndentingNewLine]", \(Table[h\_i = \[Gamma], {i, 1,
> k}];\),
>     "\[IndentingNewLine]", \(Table[h\_i = 1 - \[Gamma], {i, k + 1,
> 2}];\),
>     "\[IndentingNewLine]", \(Table[{p\_i, b\_i} = {p\_i, b\_i}, {i, 1,
> 2}];\),
>      "\[IndentingNewLine]", \(T\_p = Table[p\_i, {i, 1, 2}];\),
>     "\[IndentingNewLine]", \(T\_b = Table[b\_i, {i, 1, 2}];\),
>     "\[IndentingNewLine]", \(T\_u =
>         Table[u\_i = 1 - \[Lambda] \((b\_i - \[Beta])\)\^2 - p\_i, {i,
> 1,
>             2}];\), "\[IndentingNewLine]", \({I\_0, I\_1,
>           I\_2} = {0, \[Beta] /. \(Solve[
>                 u\_1 \[Equal] u\_2, \[Beta]]\)[\([1]\)], 1};\),
>     "\[IndentingNewLine]", \(T\_m =
>         Table[m\_i = I\_i - I\_\(i - 1\), {i, 1, 2}];\),
>     "\[IndentingNewLine]", \(T\_R =
>         Table[R\_i = \(p\_i\)
>                 m\_i - \((\(h\_i\)
>                     b\_i\^2\/2 + \((1 - h\_i)\) \((1 -
> b\_i)\)\^2\/2)\), {i,
>             1, 2}];\),
>     "\[IndentingNewLine]", \(FOCP = Table[D[R\_i, p\_i], {i, 1, 2}];\),
> 
>     "\[IndentingNewLine]", \(T\_p =
>         T\_p /. \(Solve[FOCP \[Equal] 0, T\_p]\)[\([1]\)];\),
>     "\[IndentingNewLine]", \(Table[p\_i = T\_p[\([i]\)], {i, 1, 2}];\),
> 
>     "\[IndentingNewLine]", \(FOCB = Table[D[R\_i, b\_i], {i, 1, 2}];\),
> 
>     "\[IndentingNewLine]", \(NSolve[FOCB \[Equal] 0, T\_b]\)}]\)
> 
Hi Mukhtar,

If I remember correctly the package *RealOnly* works only on 
machine-size numbers and redefine only few functions such as *Power*. 
Thus the only effect of the package on the *Solve* command is to 
suppress at the end of all computations the display of any non-real 
numbers. Internally, the algorithms still work on the complex field.

As you can see below, the imaginary part of each number is very small 
and can be dropped by using the command *Chop*. You can check that both 
solution sets are eventually identical.

In[1]:=
\!\(\*
   RowBox[{\(Clear["\<Global`*\>"]\), "\n", \(Remove["\<Global`*\>"]\),
     "\[IndentingNewLine]",
     RowBox[{
       RowBox[{
         RowBox[{
           StyleBox["{",
             FontColor->RGBColor[1, 0, 1]],
           StyleBox[\(\[Lambda], \[Gamma], k\),
             FontColor->RGBColor[1, 0, 1]],
           StyleBox["}",
             FontColor->RGBColor[1, 0, 0]]}],
         StyleBox["=",
           FontColor->RGBColor[1, 0, 0]],
         RowBox[{
           StyleBox["{",
             FontColor->RGBColor[1, 0, 0]],
           RowBox[{"1",
             StyleBox[",",
               FontColor->RGBColor[1, 0, 0]], \(2\/3\), ",", "0"}], 
"}"}]}],
       ";"}], "\[IndentingNewLine]", \(Table[h\_i = \[Gamma], {i, 1, 
k}];\),
     "\[IndentingNewLine]", \(Table[h\_i = 1 - \[Gamma], {i, k + 1, 2}];\),
     "\[IndentingNewLine]", \(Table[{p\_i, b\_i} = {p\_i, b\_i}, {i, 1, 
2}];\),
      "\[IndentingNewLine]", \(T\_p = Table[p\_i, {i, 1, 2}];\),
     "\[IndentingNewLine]", \(T\_b = Table[b\_i, {i, 1, 2}];\),
     "\[IndentingNewLine]", \(T\_u =
         Table[u\_i = 1 - \[Lambda] \((b\_i - \[Beta])\)\^2 - p\_i, {i, 1,
             2}];\), "\[IndentingNewLine]", \({I\_0, I\_1,
           I\_2} = {0, \[Beta] /. \(Solve[
                 u\_1 \[Equal] u\_2, \[Beta]]\)[\([1]\)], 1};\),
     "\[IndentingNewLine]", \(T\_m =
         Table[m\_i = I\_i - I\_\(i - 1\), {i, 1, 2}];\),
     "\[IndentingNewLine]", \(T\_R =
         Table[R\_i = \(p\_i\)
                 m\_i - \((\(h\_i\)
                     b\_i\^2\/2 + \((1 - h\_i)\) \((1 - 
b\_i)\)\^2\/2)\), {i,
             1, 2}];\),
     "\[IndentingNewLine]", \(FOCP = Table[D[R\_i, p\_i], {i, 1, 2}];\),
     "\[IndentingNewLine]", \(T\_p =
         T\_p /. \(Solve[FOCP \[Equal] 0, T\_p]\)[\([1]\)];\),
     "\[IndentingNewLine]", \(Table[p\_i = T\_p[\([i]\)], {i, 1, 2}];\),
     "\[IndentingNewLine]", \(FOCB = Table[D[R\_i, b\_i], {i, 1, 2}];\),
     "\[IndentingNewLine]", \(NSolve[FOCB \[Equal] 0, T\_b]\)}]\)
Out[17]=
\!\({{b\_1 \[Rule] 1.8560333600624024`,
       b\_2 \[Rule]
         9.206972531099032`}, {b\_1 \[Rule] \(-8.463201971504636`\),
       b\_2 \[Rule] \(-0.8172946229422916`\)}, {b\_1 \[Rule]
         0.30716861144224816`, b\_2 \[Rule] 0.9103220918432906`}}\)
In[18]:=
\!\(\*
   RowBox[{\(Clear["\<Global`*\>"]\), "\n", \(Remove["\<Global`*\>"]\),
     "\[IndentingNewLine]",
     RowBox[{
       RowBox[{
         RowBox[{
           StyleBox["{",
             FontColor->RGBColor[1, 0, 1]],
           StyleBox[\(\[Lambda], \[Gamma], k\),
             FontColor->RGBColor[1, 0, 1]],
           StyleBox["}",
             FontColor->RGBColor[1, 0, 0]]}],
         StyleBox["=",
           FontColor->RGBColor[1, 0, 0]],
         RowBox[{
           StyleBox["{",
             FontColor->RGBColor[1, 0, 0]],
           RowBox[{"1",
             StyleBox[",",
               FontColor->RGBColor[1, 0, 0]], \(2\/3\), ",", "0"}], 
"}"}]}],
       ";"}], "\[IndentingNewLine]", \(Table[h\_i = \[Gamma], {i, 1, 
k}];\),
     "\[IndentingNewLine]", \(Table[h\_i = 1 - \[Gamma], {i, k + 1, 2}];\),
     "\[IndentingNewLine]", \(Table[{p\_i, b\_i} = {p\_i, b\_i}, {i, 1, 
2}];\),
      "\[IndentingNewLine]", \(T\_p = Table[p\_i, {i, 1, 2}];\),
     "\[IndentingNewLine]", \(T\_b = Table[b\_i, {i, 1, 2}];\),
     "\[IndentingNewLine]", \(T\_u =
         Table[u\_i = 1 - \[Lambda] \((b\_i - \[Beta])\)\^2 - p\_i, {i, 1,
             2}];\), "\[IndentingNewLine]", \({I\_0, I\_1,
           I\_2} = {0, \[Beta] /. \(Solve[
                 u\_1 \[Equal] u\_2, \[Beta]]\)[\([1]\)], 1};\),
     "\[IndentingNewLine]", \(T\_m =
         Table[m\_i = I\_i - I\_\(i - 1\), {i, 1, 2}];\),
     "\[IndentingNewLine]", \(T\_R =
         Table[R\_i = \(p\_i\)
                 m\_i - \((\(h\_i\)
                     b\_i\^2\/2 + \((1 - h\_i)\) \((1 - 
b\_i)\)\^2\/2)\), {i,
             1, 2}];\),
     "\[IndentingNewLine]", \(FOCP = Table[D[R\_i, p\_i], {i, 1, 2}];\),
     "\[IndentingNewLine]", \(T\_p =
         T\_p /. \(Solve[FOCP \[Equal] 0, T\_p]\)[\([1]\)];\),
     "\[IndentingNewLine]", \(Table[p\_i = T\_p[\([i]\)], {i, 1, 2}];\),
     "\[IndentingNewLine]", \(FOCB = Table[D[R\_i, b\_i], {i, 1, 2}];\),
     "\[IndentingNewLine]", \(Solve[FOCB \[Equal] 0, T\_b]\)}]\)
Out[34]=
\!\({{b\_1 \[Rule]
         1\/820\ \((\(-1722\) -
               14661241\/\((209533 + 410\ \[ImaginaryI]\ 
\@72777)\)\^\(2/3\) +
               329294\/\((209533 + 410\ \[ImaginaryI]\ 
\@72777)\)\^\(1/3\) +
               86\ \((209533 + 410\ \[ImaginaryI]\ \@72777)\)\^\(1/3\) - \
\((209533 + 410\ \[ImaginaryI]\ \@72777)\)\^\(2/3\))\),
       b\_2 \[Rule]
         31\/10 +
           3829\/\(20\ \((209533 + 410\ \[ImaginaryI]\ 
\@72777)\)\^\(1/3\)\) +
           1\/20\ \((209533 + 410\ \[ImaginaryI]\ \@72777)\)\^\(1/3\)}, 
{b\_1 \
\[Rule] 1\/41\ \((\(-\(861\/10\)\) +
               14661241\/\(40\ \((209533 + 410\ \[ImaginaryI]\ \
\@72777)\)\^\(2/3\)\) + \(14661241\ \[ImaginaryI]\ \@3\)\/\(40\ 
\((209533 + \
410\ \[ImaginaryI]\ \@72777)\)\^\(2/3\)\) -
               164647\/\(20\ \((209533 + 410\ \[ImaginaryI]\ 
\@72777)\)\^\(1/3\
\)\) + \(164647\ \[ImaginaryI]\ \@3\)\/\(20\ \((209533 + 410\ 
\[ImaginaryI]\ \
\@72777)\)\^\(1/3\)\) -
               43\/20\ \((209533 + 410\ \[ImaginaryI]\ 
\@72777)\)\^\(1/3\) -
               43\/20\ \[ImaginaryI]\ \@3\ \((209533 + 410\ \[ImaginaryI]\ \
\@72777)\)\^\(1/3\) +
               1\/40\ \((209533 + 410\ \[ImaginaryI]\ \@72777)\)\^\(2/3\) -
               1\/40\ \[ImaginaryI]\ \@3\ \((209533 + 410\ \[ImaginaryI]\ \
\@72777)\)\^\(2/3\))\),
       b\_2 \[Rule]
         31\/10 - \(3829\ \((1 - \[ImaginaryI]\ \@3)\)\)\/\(40\ 
\((209533 + \
410\ \[ImaginaryI]\ \@72777)\)\^\(1/3\)\) -
           1\/40\ \((1 + \[ImaginaryI]\ \@3)\)\ \((209533 + 410\ 
\[ImaginaryI]\
\ \@72777)\)\^\(1/3\)}, {b\_1 \[Rule]
         1\/41\ \((\(-\(861\/10\)\) +
               14661241\/\(40\ \((209533 + 410\ \[ImaginaryI]\ \
\@72777)\)\^\(2/3\)\) - \(14661241\ \[ImaginaryI]\ \@3\)\/\(40\ 
\((209533 + \
410\ \[ImaginaryI]\ \@72777)\)\^\(2/3\)\) -
               164647\/\(20\ \((209533 + 410\ \[ImaginaryI]\ 
\@72777)\)\^\(1/3\
\)\) - \(164647\ \[ImaginaryI]\ \@3\)\/\(20\ \((209533 + 410\ 
\[ImaginaryI]\ \
\@72777)\)\^\(1/3\)\) -
               43\/20\ \((209533 + 410\ \[ImaginaryI]\ 
\@72777)\)\^\(1/3\) +
               43\/20\ \[ImaginaryI]\ \@3\ \((209533 + 410\ \[ImaginaryI]\ \
\@72777)\)\^\(1/3\) +
               1\/40\ \((209533 + 410\ \[ImaginaryI]\ \@72777)\)\^\(2/3\) +

               1\/40\ \[ImaginaryI]\ \@3\ \((209533 + 410\ \[ImaginaryI]\ \
\@72777)\)\^\(2/3\))\),
       b\_2 \[Rule]
         31\/10 - \(3829\ \((1 + \[ImaginaryI]\ \@3)\)\)\/\(40\ 
\((209533 + \
410\ \[ImaginaryI]\ \@72777)\)\^\(1/3\)\) -
           1\/40\ \((1 - \[ImaginaryI]\ \@3)\)\ \((209533 + 410\ 
\[ImaginaryI]\
\ \@72777)\)\^\(1/3\)}}\)
In[35]:=
N[%]
Out[35]=
\!\({{b\_1.` \[Rule] \(\(1.8560333600623968`\)\(\[InvisibleSpace]\)\) +
           1.109139880210888`*^-15\ \[ImaginaryI],
       b\_2.` \[Rule] \(\(9.206972531099009`\)\(\[InvisibleSpace]\)\) -
           2.7755575615628914`*^-16\ \[ImaginaryI]}, {b\_1.` \[Rule] \
\(\(0.3071686114422483`\)\(\[InvisibleSpace]\)\) +
           1.0398186376977076`*^-14\ \[ImaginaryI],
       b\_2.` \[Rule] \(\(0.9103220918432906`\)\(\[InvisibleSpace]\)\) +
           1.7763568394002505`*^-15\ \[ImaginaryI]}, {b\_1.` \[Rule] \
\(-8.463201971504644`\) - 1.0398186376977076`*^-14\ \[ImaginaryI],
       b\_2.` \[Rule] \(-0.8172946229422982`\) -
           1.7763568394002505`*^-15\ \[ImaginaryI]}}\)
In[36]:=
Chop[%]
Out[36]=
\!\({{b\_1.` \[Rule] 1.8560333600623968`,
       b\_2.` \[Rule] 9.206972531099009`}, {b\_1.` \[Rule]
         0.3071686114422483`,
       b\_2.` \[Rule]
         0.9103220918432906`}, {b\_1.` \[Rule] \(-8.463201971504644`\),
       b\_2.` \[Rule] \(-0.8172946229422982`\)}}\)

Best regards,
/J.M.