Political forecasting

An extrapolation is an estimated extrapolation an overall score of a partial result. It is applied when not all the information for the overall result, the amount of information is too large to cope with reasonable effort or a full survey is not possible. To provide the most accurate extrapolation of the results from use must take into account all possible aspects and in this case be numerically large enough. One then speaks of a representative sample or a sample.

Examples

• In elections, for example, election research institutes from the first election results trying to predict the end result. Delineate from the selection extrapolation is the choice prognosis. Here, in addition to the previously counted votes, the results of the survey of different groups of people ( for example, the unemployed, parents, singles, retirees) are taken into account according to their number in the electoral area in question. Targeted is thereby questioned in the constituencies whose result was particularly close to the official results at the vorgegangenen choice.

Other examples of the use of extrapolation are the basis

• The partial counts of the components of blood for diagnostic support or
• The counts of different areas of our or another galaxy to guesstimation determining the total number of the constituent stars.

Bound and free extrapolation

Free extrapolation: Under general and classical understanding extrapolations are three set tasks. Example: 10 students interviewed had a total pocket money ( sum over all 10) from 100 €. How much total pocket money, the entire class of 30 students? Here are regarded as estimators of a characteristic X ( eg for the total value ). It thereby only the observed feature values ​​, the sample size was (possibly in separate layers ) and the scope of the whole (possibly in separate layers ) is used.

Bound extrapolation: The extrapolation bound prior information can be used. If one uses external information, one also speaks of bound extrapolation ( engl. "using auxiliary information at the estimation stage" ).

Historically, it has differences estimator here ( variable in Y -intercept) and ratio estimators used ( variable in the slope ). Today, we used regression estimators which combine both properties. In small samples, however, regression estimators are more susceptible to outliers.