O
Oliver Heaviside
regardez vous sil vou plait this result of multiplying a7by7 matrix by its
MINVERSE "inverse":
1.0E+00 3.7E-08 2.3E-06 1.4E-04 8.2E-03 4.9E-01 2.9E+01
1.0E-10 1.0E+00 2.3E-07 1.1E-05 5.9E-04 3.5E-02 1.6E+00
2.4E-11 1.3E-09 1.0E+00 3.7E-06 2.1E-04 1.1E-02 6.6E-01
-8.4E-13 -4.3E-11 -2.4E-09 1.0E+00 -7.5E-06 -3.8E-04 -2.2E-02
1.2E-14 5.4E-13 3.3E-11 1.5E-09 1.0E+00 4.4E-06 2.7E-04
-9.7E-17 -4.7E-15 -2.8E-13 -1.3E-11 -7.0E-10 1.0E+00 -1.8E-06
4.2E-19 2.0E-17 1.3E-15 6.0E-14 3.2E-12 1.3E-10 1.0E+00
see the top right hand term 29!!! the curve fitted from this does not fit!
in the help at
http://office.microsoft.com/en-us/excel/HP052091791033.aspx?pid=CH062528291033
it says of MINVERSE
The size of the array must not exceed 52 columns by 52 rows. If it does, the
function returns a #VALUE! error.
MDETERM gives -3.25223E+110
for this 10 by 10 matrix, the first 7by7 of which was referred to above
9.0E+00 2.8E+02 1.2E+04 5.5E+05 2.7E+07 1.4E+09 7.5E+10 4.1E+12 2.2E+14 1.3E+1
2.8E+02 1.2E+04 5.5E+05 2.7E+07 1.4E+09 7.5E+10 4.1E+12 2.2E+14 1.3E+16 7.0E+1
1.2E+04 5.5E+05 2.7E+07 1.4E+09 7.5E+10 4.1E+12 2.2E+14 1.3E+16 7.0E+17 4.0E+1
5.5E+05 2.7E+07 1.4E+09 7.5E+10 4.1E+12 2.2E+14 1.3E+16 7.0E+17 4.0E+19 2.3E+2
2.7E+07 1.4E+09 7.5E+10 4.1E+12 2.2E+14 1.3E+16 7.0E+17 4.0E+19 2.3E+21 1.3E+2
1.4E+09 7.5E+10 4.1E+12 2.2E+14 1.3E+16 7.0E+17 4.0E+19 2.3E+21 1.3E+23 7.5E+2
7.5E+10 4.1E+12 2.2E+14 1.3E+16 7.0E+17 4.0E+19 2.3E+21 1.3E+23 7.5E+24 4.3E+2
4.1E+12 2.2E+14 1.3E+16 7.0E+17 4.0E+19 2.3E+21 1.3E+23 7.5E+24 4.3E+26 2.5E+2
2.2E+14 1.3E+16 7.0E+17 4.0E+19 2.3E+21 1.3E+23 7.5E+24 4.3E+26 2.5E+28 1.4E+3
1.3E+16 7.0E+17 4.0E+19 2.3E+21 1.3E+23 7.5E+24 4.3E+26 2.5E+28 1.4E+30 8.4E+31
is MINVERSE working as intended, not with 52 by 52 but 7 by 7?
is this why the polynomial fitted to xy plots is 6th order no bigger?
can anyone advise how to invert nonsingular matrices, up to 52 by 52,
eg 10 by 10 would be good, so the product of the matrix and its
inverse looks just like this?
1000000
0100000
0010000
0001000
0000100
0000010
0000001
and no 29s or such anywhere off diagonal please?
thanks
jim in sydney
MINVERSE "inverse":
1.0E+00 3.7E-08 2.3E-06 1.4E-04 8.2E-03 4.9E-01 2.9E+01
1.0E-10 1.0E+00 2.3E-07 1.1E-05 5.9E-04 3.5E-02 1.6E+00
2.4E-11 1.3E-09 1.0E+00 3.7E-06 2.1E-04 1.1E-02 6.6E-01
-8.4E-13 -4.3E-11 -2.4E-09 1.0E+00 -7.5E-06 -3.8E-04 -2.2E-02
1.2E-14 5.4E-13 3.3E-11 1.5E-09 1.0E+00 4.4E-06 2.7E-04
-9.7E-17 -4.7E-15 -2.8E-13 -1.3E-11 -7.0E-10 1.0E+00 -1.8E-06
4.2E-19 2.0E-17 1.3E-15 6.0E-14 3.2E-12 1.3E-10 1.0E+00
see the top right hand term 29!!! the curve fitted from this does not fit!
in the help at
http://office.microsoft.com/en-us/excel/HP052091791033.aspx?pid=CH062528291033
it says of MINVERSE
The size of the array must not exceed 52 columns by 52 rows. If it does, the
function returns a #VALUE! error.
MDETERM gives -3.25223E+110
for this 10 by 10 matrix, the first 7by7 of which was referred to above
9.0E+00 2.8E+02 1.2E+04 5.5E+05 2.7E+07 1.4E+09 7.5E+10 4.1E+12 2.2E+14 1.3E+1
2.8E+02 1.2E+04 5.5E+05 2.7E+07 1.4E+09 7.5E+10 4.1E+12 2.2E+14 1.3E+16 7.0E+1
1.2E+04 5.5E+05 2.7E+07 1.4E+09 7.5E+10 4.1E+12 2.2E+14 1.3E+16 7.0E+17 4.0E+1
5.5E+05 2.7E+07 1.4E+09 7.5E+10 4.1E+12 2.2E+14 1.3E+16 7.0E+17 4.0E+19 2.3E+2
2.7E+07 1.4E+09 7.5E+10 4.1E+12 2.2E+14 1.3E+16 7.0E+17 4.0E+19 2.3E+21 1.3E+2
1.4E+09 7.5E+10 4.1E+12 2.2E+14 1.3E+16 7.0E+17 4.0E+19 2.3E+21 1.3E+23 7.5E+2
7.5E+10 4.1E+12 2.2E+14 1.3E+16 7.0E+17 4.0E+19 2.3E+21 1.3E+23 7.5E+24 4.3E+2
4.1E+12 2.2E+14 1.3E+16 7.0E+17 4.0E+19 2.3E+21 1.3E+23 7.5E+24 4.3E+26 2.5E+2
2.2E+14 1.3E+16 7.0E+17 4.0E+19 2.3E+21 1.3E+23 7.5E+24 4.3E+26 2.5E+28 1.4E+3
1.3E+16 7.0E+17 4.0E+19 2.3E+21 1.3E+23 7.5E+24 4.3E+26 2.5E+28 1.4E+30 8.4E+31
is MINVERSE working as intended, not with 52 by 52 but 7 by 7?
is this why the polynomial fitted to xy plots is 6th order no bigger?
can anyone advise how to invert nonsingular matrices, up to 52 by 52,
eg 10 by 10 would be good, so the product of the matrix and its
inverse looks just like this?
1000000
0100000
0010000
0001000
0000100
0000010
0000001
and no 29s or such anywhere off diagonal please?
thanks
jim in sydney