1 / 21

Sampling for Surveys in the Dutch Statistical Bureau

Sampling for Surveys in the Dutch Statistical Bureau. Ida H. Stamhuis One of the chapters of the book The Statistical Mind in the Netherlands 1850-1940 (to be published in 2008) (co-authors Jelke van Bethlehem, Jacques van Maarseveen) For ‘the’ story I refer to that chapter

thuy
Télécharger la présentation

Sampling for Surveys in the Dutch Statistical Bureau

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Sampling for Surveys in the Dutch Statistical Bureau Ida H. Stamhuis One of the chapters of the book The Statistical Mind in the Netherlands 1850-1940 (to be published in 2008) (co-authors Jelke van Bethlehem, Jacques van Maarseveen) For ‘the’ story I refer to that chapter Now I will suggest a more analytical approach Woudschoten, 28 and 29 September 2007

  2. Continuation of my Woudschoten paper of 2005 Segmentation of various forms of statistics • Segmentation between ‘secret’ statistical practice of the public administration and statistical theory • Socio-economic-political explanations (didn’t go into that) • Segmentation within theoretical statistics • Two different forms of knowledge • Carried out by people with different ways of thinking • Carried out by people belonging to different intellectual cultures • No real communication between them

  3. Lobatto’ statistics: ‘quantitative-probabilistic’

  4. Vissering’s statistics representative of that of the ‘Statistical Movement’: qualitative-quantitative • Statistics as systematic description of all aspects of state / society • Statistics new discipline at the faculties of law

  5. From the viewpoint of circulation of knowledge • Conclusion: segmentation; no circulation (for the time being) • It asks for an explanation. • In 2005 I concluded with the question:What kind of expertise is necessary? Result was discouraging • Another approach: Let us try out a theoretical social scientific notion and look at a related case in which the border has ultimately passed: ‘habitus’ of Bourdieu • Social experiences and circumstances result in ‘mental structure’ through which the world is experienced: habitus • Consequence: not easily changed by intellectual arguments, but rather by change of experiences and circumstances

  6. Survey: observational study of social or economic factors of populations, emerged from the ‘Vissering’ statistical world • A sample is “a part for the whole”; a sample must be ‘representative’

  7. 19th century ‘monograph studies’ considered ‘representative’ • ‘typical entities’ • No exceptional cases • The whole is more than the sum of the parts

  8. Kiaer: purposive samples representative • In purposive sampling: variation also taken into account • Around 1900 discusssion about the choice between complete enumeration and purposive sampling in the international statistical community: • Von Mayr: “One cannot replace by calculation the real observation of facts”

  9. In England development of mathematically oriented (bio)statistics including random sampling; Pearson and Fisher • In random sampling expertise of the statistician is replaced by ‘blind’ chance • 1906 Bowley proposed random sampling in the English statistical community

  10. 1924 Commission in the International Statistical Community: report • Random and purposive selection both reason to exist • Describe meticulously the sampling procedure in each investigation • ‘run with the hare and hunt with the hounds’

  11. Prominent Dutch Statisticians involved: • Verrijn Stuart: “I state that in principle no representative method, one or another, can have the significance of a complete enumeration of the phenomenon of study” • Otherwise random selection because of the ‘Law of Large Numbers’ • Methorst: purposive sample “helps to save a great deal of expense and labor”

  12. Surveys mainly executed by official statisticians: educated in law, • directors of the CBS (Verrijn Stuart, Methorst, Idenburg), Van Zanten (director Amsterdam SB)

  13. Dutch Statistical textbooks • 1910 Verrijn Stuart: • Monograph method • Purposive selection • 1927 Van Zanten: Law of Large Numbers • Purposive selection • Random selection • No preference between the two methods

  14. Dutch Statistical textbooks • From the 1930’s people with more exact background invaded the Dutch statistical community: Accountants Bakker and Stridiron authors of textbooks, Tinbergen, Holwerda (actuary) • Bakker 1934: random sampling, probability theory in words. • 1939-2: In Foreword thanks to Methorst as well as Tinbergen and Holwerda • 1941-3: distinguishes between quality control and surveys • In distinct section Derksen introduces mathematical formula • Section about opinion polls

  15. Dutch Statistical textbooks • Tinbergen 1936 Grondproblemen der Theoretische Statistiek (Basic Problems of Theoretical Statistics) • Random samples obvious • Accountant Stridiron 1941 Handboek der Bedrijfseconomische Statistiek (Handbook of Statistics of Business Economics): Mathematical parts written by Tinbergen • Random samples obvious • 1943-2 Derksen Co-author chapter ‘Samples”: Survey as well as Quality Control

  16. Practice in CBS • Methorst tested in 1924 ‘representative method’ , partly purposive, partly random: it did not work • Then surveys for economic indicator’s like consumer price index; national expenditure surveys necessary: recruitment of households through labor unions and advertisements • Etc. Samples often purposive, sometimes mixed with random

  17. Sampling in Market Research • Various organizations • Unilever established Lintas (1934) and IHO (1938) • NSS (1940) • NIPO (1945) • Sampling obvious: CBS director Idenburg felt in 1948 the need to explain why CBS chose for complete census • Sampling method not always clear • Problem of Non-response more attention than sampling method

  18. Conclusions • Conceptual hurdles to go from completeness to samples: information is lost • Conceptual hurdles to go from purposive to random sampling: opinion of expert not relevant; mathematical principles not clear to the powerful statisticians in the first half of the twentieth century

  19. Conclusions • Other hurdles: (“Socio-economic-political”) • Automation makes completeness cheap • Administration doesn’t like to fire officials • Complete material available • Organization structures are not easily changed

  20. Why ultimately randomness accepted? • Not because statisticians start to understand the mathematical background • Rather because they become convinced of its relevance • Intermediaries (belong to both communities) play an important role • Other hurdles disappear

  21. Habitus of Bourdieu?

More Related