Summary
To estimate is to approximate. Since, for operational or conceptual reasons, not all the information is available, some characteristics of the variables will inevitably be unknown; these can be any of the measures of location, dispersion or association seen earlier. Their true value is what is called a parameter.
The goal is precisely to approximate these parameters through different formulas, called estimators. For each case there is no single possible estimator, although some meet certain properties that make them optimal, or at least better than others.
Data are needed to reach an approximate value of the parameters through the estimators. This is where the sampling process comes in: obtaining data, necessarily at random, so that the estimation yields valid results. The larger and more representative a sample is with respect to the population, the closer the values of the estimators get to the values of the parameters, always differing from one sample to the next but with less variation as the sample size approaches the full number of units of analysis (or infinity).
The variation in the results of the estimators plays a fundamental role in managing uncertainty. It is common not to state a single value for the estimator but rather an interval. It can then be said that this interval, with a certain degree of confidence, will contain the true value of the parameter a given number of times (number of samples or experiments).
All the results obtained are subject to the assumption that the population (of units of analysis) follows a given distribution. When trying to estimate the mean, the variance or a proportion, it must be kept in mind that these are precisely parameters used in particular distributions, some of those seen earlier, such as the normal or the binomial. This, in a way, makes the calculations easier and gives greater certainty about the conclusions; but in many cases the distributions are not known, or some of these parameters are not the target of estimation. In that case, nonparametric estimation methods are needed, which are covered later.
Further reading
- Estimación y parámetros (in Spanish). Kelmansky, Estadística para todos, p. 55.
- Variabilidad entre muestra y muestra (in Spanish). Kelmansky, Estadística para todos, pp. 58–63.
- Teorema central del límite (in Spanish). Kelmansky, Estadística para todos, pp. 200–207.
- Distribuciones de muestreo de estadísticas. Canavos, Probabilidad y estadística. Aplicaciones y métodos (in Spanish), pp. 218–238.
- Estimación por intervalos (in Spanish). Kelmansky, Estadística para todos, pp. 214–228.
- Estimación e intervalos de confianza. Lind, Marchal and Wathen, Estadística aplicada a los negocios y la economía (in Spanish), 15th ed., pp. 297–315.
- Propiedades deseables de los estimadores puntuales. Canavos, Probabilidad y estadística. Aplicaciones y métodos (in Spanish), pp. 251–261.
- Confidence Intervals. STAT 500: Applied Statistics, Penn State, lesson 5.