Analizy weryfikacyjne w badaniach z zastosowaniem wywiadu kwestionariuszowego, ich rodzaje i możliwości
Streszczenie
Application of standardized, techniques of data gathering permitted to obtain quantitative research results. However, a researcher cannot evaluate precisely the value of these results. He can only — but not always — assess the magnitude of the sampling error without being able to estimate errors of a different origin and especially response errors. Elaboration of methods of estimating such errors should be one of the purposes of methodological investigations.
Analyses of surveys using questionnaire interviews has shown that, in evaluating their results, it is not possible to employ the methods of evaluation developed in other sciences. This evaluation, most often imprecise, is the purpose of verificational analyses conducted in sociology. Two principal kinds of such analyses may be distinguished: external and internal one. External analysis consists in checking the data obtained by means of questionnaire interviews against the data obtained as a result of other research processes. The latter data may relate 1° to the members of an identical collectivity, 2° or to a comparable collectivity (most frequently a split-ballot technique is used). In the first case, the evaluation may be performed not only on the basis of a comparison between the data describing the same phenomenon, but also 3° on the basis of other data, relating to other phenomena strongly correlated with those being the object of the research. These three types of external analyses may be further divided according to whether the control information has been obtained by means of a procedure which is absolutely trustworthy, only more trustworthy, or as trustworthy as the information whose trustworthiness is to be evaluated. In consequence 9 specific types of verificational analyses are obtained. Their possibilities as well as the scope of the application are diversified. It is only those analyses in which the same information is obtained twice from the identical collectivity, and in which absolutely trustworthy data are used, that permit one ascertain which answers turned out to be valid, i.e. contain a true information, and, as a result, to assess the magnitude of the gross error as well as the net error. The magnitude of the net error can be assessed also in case the control results have been obtained through absolutely trustworthy research procedure applied to another, comparable collectivity of respondents. However the application of the both types of analyses is limited due to the lack of absolutely trustworthy procedures in investigations of most social phenomena. Other types of analyses do not show this limitation but they do not yield trustworthy information about the errors or any quantitative results at all. The article discusses in detail the possibilities of each of the distinguished 9 types of analyses.
Internal verification consist in an analysis and evaluation of the processes that that yield the information desired. If a given process goes on appropriately, then the respondent’s answer is also appropriate and should be valid. In order to be able to evaluate every information-getting process, it is necessary to compared it with a model of an appropriate course of such a process. The article presents the method of construction of such models, with regard to every questionnaire question and the information desired. The method is based on the scheme of inference from the indicators about phenomenon under study, the scheme having been adapted to the questionnaire interview. The first premise, in this inference, relates to the asking a respondent a given question as well as to the kind of an answer given by him to that question; the second premise, concerns the relation between that answer and the statement about the research unit, the statement concerning the information desired. This scheme constitutes the most general form of the model. In order to make it more specific, one looks for these statements concerning the information-getting process, which warrant the truthfulness of the premises. If the statements are true, then the premises will be also true as well as the conclusion drawn from them and, in consequence, the information obtained. The total truthfulness of these statements constitutes then a sufficient condition of truthfulness of that information. The statements concern: 1° external manifestations of the process through which the information is formed (asking a question, recording and classification of the answer); 2° correctness and consistency of the classification used; 3° internal reactions of a respondent to the questionnaire question; 4° eventually, the relation between the interview situation and the real-life situations; 5° sometimes, the regularities in the occurrence of certain phenomena outside the interview situation. Only the statements 1° and 2° are always identical, regardless of the type of a question, the remaining statements vary. The model constructed this way may be also used in an analysis of the correctness of a questionnaire question. The data concerning the internal reactions of a respondent can be obtained by means of a specially designed interview about an interview which yields the information about how a respondent has understood the question and formulated an answer. Psychological tests may be also used for this purpose. The result of internal verification does not necessarily overlap with that obtained by external verification. If, by means of the latter, it is ascertained that a given answer is valid, this does not necessarily mean that the course of the process by which that answer has been obtained, was appropriate. The answer may be valid accidentally.
The use of an internal verification is limited, due to its labouriousness; it can be however used for the verification of those answers and the corresponding information which cannot be verified externally, e.g. those concerning opinions.
It is recommended to elaborate verificational procedures lending themselves to practical application and permitting to assess the magnitude of errors. Such procedures should consist of a combination of various types of analyses.
Collections