Invert Matrix Command

A and B are matrices, which A is invers from B. Is there any differences between A=B^-1; A=B.\1; A=inv(B);

coz i have different result from 3 command above

[Jan]

What about trying it in Matlab and comparing the results?

[andy ganteng]

Yes i have tried, but in my case when i used A=B^-1 and A=inv(B) i get Inf result. it was different with A=B.\1 , although explanation in matlab help showed simmiliarity between all those command, that was i guess....maybe im wrong.

[Grzegorz Knor]

As I mentioned, "A=B.\1" method is wrong. First, check the determinant of the matrix. If its determinant is 0, the matrix is singular

[andy ganteng]

is there any method to avoid Inf result when create inverse matrix?

[Jan]

"Avoid"? Building the inverse is a well-defined mathematical operation. There are no degrees of freedom which allow to modify the result.

It is like a division by zero: You can catch this case, but you cannot avoid, that the result is Inf (or NaN for some cases).

[andy ganteng]

what about pinv?is that another choice for me?

[Grzegorz Knor]

Can you write how you make the matrix B and why do you need the inverse matrix?

[andy ganteng]

This is my data, i call it matrix y, which consist y1 and y2.

y1=y(:,1); and y2=y(:,2);

91.3600000000000 9.09000000000000

81.9800000000000 8.04000000000000

83.3900000000000 7.95000000000000

89.9300000000000 7.32000000000000

83.3800000000000 7.94000000000000

88.2100000000000 7.26000000000000

83.1100000000000 6.88000000000000

79.0700000000000 6.44000000000000

94.0500000000000 10.5400000000000

87.3666666666667 8.62756395875771

51.6500000000000 3.79187935524974

96.3123697889276 9.05000000000000

83.5201129290811 6.47628937657751

88.2822635556373 6.53120884650403

97.7485157937070 9.26010524849543

62.9108356860225 6.21188000000000

86.8054178430113 6.10960000000000

87.2876058710136 6.70809271547082

31.7541784301126 3.10103746017809

31.3541784301126 3.89263270542303

31.0704319591464 3.93979332207437

31.8054178430113 2.41932709948679

31.7571174473266 2.45365318072931

32.8654584109626 2.94404622817266

87.1108356860225 6.41229552674817

91.1162535290338 7.32147940127193

76.8792490107883 6.29018684796355

95.7053482338574 7.97101408843632

64.1054178430113 4.45535798833371

30.5254178430113 2.78703470631920

32.6825070580676 3.33189666711284

32.1254178430113 2.89685219459403

32.7709486151037 2.72341621277990

32.1054178430113 2.79210000000000

32.9850988129460 3.42999394839633

99.1000000000000 10.8835853352147

then a wanna make covariance matrix from y1 and y2. After that i need the inverse from that covariance matrix.

[andy ganteng]

I used this command

y1=y(:,1);

y2=y(:,2);

nn=length(y1);

vary1=((y1-mean(y1))*(y1-mean(y1))')/(nn);

vary2=((y2-mean(y2))*(y2-mean(y2))')/(nn);

covy12=((y1-mean(y1))*(y2-mean(y2))')/(nn);

covy21=((y2-mean(y2))*(y1-mean(y1))')/(nn);

w=[vary1 covy12;covy21 vary2];

W=inv(w);

the result always Inf.

[andy ganteng]

I used W for the weight matrix in weighted least square regression for biresponse regression. please help.

[Grzegorz Knor]

Each row in a matrix w is linearly dependent, so rank(w) = 1.

hold all

for k=2:72

plot(w(1,:)./w(k,:))

end

There is no an inverse matrix in this case.

[andy ganteng]

thanks

Answer 1

[Grzegorz Knor]

Statement A=B.\1 doesn't calculate an inverse matrix. You probably mean:

A2=B\eye(size(B))

B^-1 and inv(B) give the same result. B\eye(size(B)) is a little different, because it uses different algorithm.