You are on page 1of 2

Efficiency

Efficiency, like consistency, it is a large sample property. The efficiency of an estimator is


usually stated in relative terms.

Efficiency is a property by which we compare an estimator with other one on the bases of its
variances. Sometimes the mean square error (MSE) is used as the replacement of the variance
for possible comparison of the estimator.

Efficient estimator
The two unbiased statistics are used as the estimator for some parameter the one whose
sampling distribution has the smallest variance is more efficient estimator than the other.

Let we have two estimators 𝜃̂1 and 𝜃̂2 of the same parameter 𝜃. Then 𝜃̂1 will be said more
efficient estimator than 𝜃̂2 , if and only if

𝑉𝑎𝑟(𝜃̂1 ) < 𝑉𝑎𝑟(𝜃̂2 )

𝑉𝑎𝑟(𝜃̂1 )
< 1.
𝑉𝑎𝑟(𝜃̂2 )

OR

The relative efficiency of 𝜃̂1 to 𝜃̂2 is given by

𝑉𝑎𝑟(𝜃̂2 )
𝐸𝑓𝑓𝑖𝑐𝑖𝑒𝑛𝑐𝑦 = .
𝑉𝑎𝑟(𝜃̂1 )

If 𝐸𝑓𝑓𝑖𝑐𝑖𝑒𝑛𝑐𝑦 > 1 the 𝜃̂1 is an efficient estimator

If 𝐸𝑓𝑓𝑖𝑐𝑖𝑒𝑛𝑐𝑦 < 1 the 𝜃̂2 is an efficient estimator

For example: both sample mean and median are unbiased and consistent estimators (But we
have to check which one is more efficient)

𝑆𝑎𝑚𝑝𝑙𝑒 𝑀𝑒𝑎𝑛 = 𝑋̅

𝑆𝑎𝑚𝑝𝑙𝑒 𝑀𝑒𝑑𝑖𝑎𝑛 = 𝑋̃

As

𝜎2
𝑉𝑎𝑟(𝑋̅) =
𝑛
𝜋 𝜎2
𝑉𝑎𝑟(𝑋̃ ) = ( )
2 𝑛

Now

𝜋 𝜎2
𝑉𝑎𝑟(𝑋 ) 2 ( 𝑛 ) 𝜋
̃
𝐸𝑓𝑓𝑖𝑐𝑒𝑛𝑐𝑦 = = = = 1.57
𝑉𝑎𝑟(𝑋̅) 𝜎2 2
𝑛
As, 𝐸𝑓𝑓𝑖𝑐𝑖𝑒𝑛𝑐𝑦 > 1 so (𝑋̅) is more efficient estimator than median.

Note: The efficiency of sample mean relative to sample median is 1.57 or 157% which means
that a sample mean calculated from sample of size 100 can do the same job as the sample
median calculated from a sample of size 157.

Alternatively, the efficiency of the sample median is


𝜎2
𝑉𝑎𝑟(𝑋̅) 𝑛 2
𝐸𝑓𝑓𝑖𝑐𝑒𝑛𝑐𝑦 = = 2 = = 0.637
𝑉𝑎𝑟(𝑋̃ ) 𝜋 (𝜎 ) 𝜋
2 𝑛

This means that the precision of the estimator from median of a sample of 1000 is about the
same as that for the mean of a sample 637.

You might also like