What do you mean "当一元线性回归方程是经过原点的模型时"?

If your data is generated by a model,

y=a+b*x + err   err ~ normally distributed  a and b are not zero.

When you force the regression line go through the origin, then the distribution of sum of errors is not known.

过原点和不过原点的自由度应该是一样的