y = a * x^4 + b * x^3 + c * x^2 + d * x + e，

If each group (x, y) is substituted into the above formula, in theory, most 5 yuan equations of 1 degree can be obtained (if the coefficient matrix is not linearly correlated), and a to e, absolutely accurate x, y can be obtained by solving them.

Instead of using the least square method,

The correlation coefficient and correlation test in the problem are linear regression by least square method.

y = a * x + b

When, you can use software to calculate parameters, which is useless.

I haven't heard of the sum of squares of deviation, and I have heard of correcting the variance of deviation. It's useless.

Supplement: this method in the fourth layer is solved by adjoint matrix, which is difficult, but it is not actually used.

If you use MATLAB, you can directly find the inverse matrix, such as

[ A B C D ] [ W ]

A=[ E F G H ] b = [ W X Y Z ]' = [ X ]

[UK]

[ M N O P ]

(Note the transposition of the first one in B)

Then use inv(A)*b or A\b to get the b 1 to b4 you need.