Point estimation
An essential task of inferential statistics is the most accurate determination of certain qualities or quantities of a population. Since the observations are assumed to be random, the determination may not be accurate, but is done as an estimate. A point estimate is in this sense a function that assigns to present observations an estimate of the size of interest. In most applications the quantity of interest is a parameter of the probability distribution of observations (for example, the average of a normal distribution ). The point estimate is different from the interval estimation substantially by the fact that its result is a single value and thus no information about the accuracy of the estimate can be derived from it. To form interval estimates point estimates are used as a basis.
A point estimate is a function of random observations, a point estimate of the calculated value of the point estimate for the present observations.
Point estimator are particularly used for the estimation of:
- Mean values
- Variances
- Standard deviations
The quality of a point estimator is assessed on the basis of various properties. The central properties in the literature are:
- Unbiasedness ( unbiasedness, genuineness ): A point estimator is unbiased if it correctly indicates the average of the actual value of the interest size. In this sense, the estimate does not have a systematic error.
- Consistency: Consistency means clear that the point estimate tends to be closer to the actual value of the interest size for a growing number of observations.
- Efficiency: efficiency is a point estimator if its spread in comparison with other point estimates is minimized. In this sense, an efficient estimator has no unnecessary scattering.
Example: There are random observations, all of which are independent and normally distributed with unknown mean and variance. The average value can be estimated by the arithmetic mean of the observations:
The expected value of this point estimate is the true parameter value:
Therefore, the estimator is unbiased.