Question

If T(x)be an estimator of θ , then bias term can be defined as,​

Answer 1

the bias term is define as the difference between it's expectation and true value.

Answer 2

Bias term can be defined as  $bias(\hat\theta) = E_\theta(\hat\theta) -\theta$.

Explanation:

• Suppose that $X_1, . . . , X_n$ are iid, each with pdf/pmf $f_X (x | \theta), \theta$ unknown.
• We aim to estimate $\theta$ by a statistic, i.e by a function T of the data.
• If $X = x = (x_1, . . . , x_n)$ then our estimate is $\hat{\theta} = T(x)$ (does not involve $\theta$).   Here,  $T(X)$  is our estimator of $\theta$, and is a random variable since it inherits random fluctuations from those of X.    
• Now we have $\hat{\theta} = T(x)$  is an estimator of $\theta$.
• Then the bias of $\hat\theta$ is the difference between its expectation and the ’true’ value:           $bias(\hat\theta) = E_\theta(\hat\theta) -\theta$