577.1
The Indicator of External Validity As Aggregate Criterion Quality of Survey
The Indicator of External Validity As Aggregate Criterion Quality of Survey
Thursday, July 17, 2014: 8:30 AM
Room: 416
Oral Presentation
The external validity criteria in sociological survey methodology and practice are used rarely. The exceptions are electoral polls. The well known Gallup and others polls obtained the trust of public and professionals due to external (predictive) validity, although the term had not yet been coined. The predictive validity is the simple and most convincing proof of the high quality of all elements and stages of the survey.But the range of sociological topics allowing simple criteria of the predictive validity is limited. Methods for assessing the external validity developed in psychology, medicine and experimental economics where in many research single target parameter and one external parameter analogue of an unbiased independent source are used. Therefore, in the most sociological survey various complex internal validity criteria are used for evaluation of survey quality. These methods allow to evaluate the common error for all variables of a survey under the condition that the sample is random. However the last few decades a number of difficulties with probability sampling have been cropping up in the expanding range of survey topics. The basis of application of the internal validity methodology has been shrinking too. The paper offers a few steps expand the application of criteria of external validity: – install for surveys on specific topics limit application of the criteria of internal validity - “unamendable nonresponse rate” (nonresponse is the main factor that makes a non-random sample); – develop and apply external validity indicators for assessing the quality of a survey on selected topics - "item external validity"; – include ancillary questions in the questionnaire, which will compare the survey data with data from independent external sources. In our surveys of social settings in Saint-Petersburg (2007-11) we used as a measure of external validity and simple distribution, and complex calculated indicator of latent social phenomena.