Quantification, counting or simplified enumeration, much more than a simple method, the survey is known as first-rate tools allowing the establishment of a representative sample of a statistical population (parent group). The main purpose is to extrapolate with a margin of error to minimize, an information (an estimator) on the entire population. The reasons for the choice of the use of this technique are various, among which, the cost, the duration, the difficulties of a census, the budget, the staff ... etc. Maintaining a subset of a universe to globalize its information considers several criteria including efficiency, convergence, consistency (or robustness), Fischer information, the accuracy of the estimator obtained ... etc. On the other hand, the most important of these requirements on an estimator is precision (its convergence). In general, the survey technique is divided into two sections: on the one hand, the empirical survey (non-random) and on the other hand, the random survey. The core of the survey focuses on the notion of representativeness, which was first considered by George Gallup founder of the American Institute of Public Opinion. Therefore, the implementation must be done with more care and in the strictest possible rigor and consistent with the principles of sounding. Probabilistic methods refer to the assignment of a probability to all members of the population and to possible samples, known as inclusion probability (or survey weight). Among these are a variety of method choices based on cost, simplicity of sampling technique, structure, character dispersion within a parent population. The simplest method in probabilistic sampling technique is that of simple random draw with discount and without discount. The choice between them depends on the size of the sample. Generally, if the size of it is large, the technique with discount is more appropriate. Nevertheless, the precision of the estimator is obtained with the one not delivered. When one wishes to select by group stratification of homogeneous strata between them (intra-low variance), one also realizes that the estimator is precise that it is not. Improving quality by reducing dispersion is the purpose of sampling. To do this, we stratify according to the variable of interest. Indeed, in each stratum considered, the random selection of individuals with or without discount is maintained. From this point of view, the sampling runs in two phases. Until now, we have seen that the no-discount is preferred to the one with discount in the case of small sample or finite size. Stratification is also an added value in the accuracy of the estimator. The singularity between these methods, the probability of selection (or draw) is identical for all individuals and well known in advance. The procedure of the systematic survey or that of fish sampling or Bernoulli looks a little different, the probability of inclusion or weight of the survey is unequally distributed in the population. It is on the mathematical level, a function of auxiliary information. A variable correlated with the variable of interest. The ideal is always shot at probability with discount or without discount. Beyond the simple sounding (elementary), for the complex case namely that by cluster and with one or more degrees. The advantage lies in reducing the cost and improving the accuracy and therefore the quality of the estimator. These sampling techniques are used in the case where the enumeration area, the building blocks are homogeneous with each other (low intergroup variance). Its implementation also depends on the field of study, the structure, the variability and especially the availability of an exhaustive list of individuals. In fact, elementary sampling is also involved in these processes and is one of the last phases of a multi-stage survey. As far as empirical methods are concerned, the quota method (pseudo-random) is generally used. The selection (or draw) is done in such a way that the sample can have the same characteristic as the population. Not knowing a priori, the probability of inclusion, extrapolation or inference must be taken with more precaution. The consideration of this method lies in the fact of dedicating a quota for each characteristic on which the sample survey is conducted. The other empirical methods used in case no population information is available (base case unavailable) are convenience (or voluntary) sampling; Snowball; a priori ... etc.
Abdi-Basid ADAN, 2018.
© 2026 The Abdi-Basid Courses Institute (tABCi). All rights reserved.
馃實 The EcoClimate Hub | Focus: Climate Science & Environment.
Contact : Abdi-Basid ADAN [abdi-basid@outlook.com]
© 2017-2026 The Abdi-Basid Courses Institute (tABCi). All rights reserved.
