+ Reply to Thread
Page 2 of 3 FirstFirst 1 2 3 LastLast
Results 16 to 30 of 44

Thread: [PoSt] Philsophy of Statistics

  1. #16
    Admin
    Points: 15,849, Level: 80
    Level completed: 99%, Points required for next Level: 1
    quark's Avatar
    Location
    Canada
    Posts
    469
    Thanks
    23
    Thanked 163 Times in 65 Posts

    Re: [PoSt] Philsophy of Statistics




    This article from StatsBlogs seems to be relevant.

    E.S. Pearson’s Statistical Philosophy
    http://www.statsblogs.com/2012/08/16...al-philosophy/

  2. #17
    Points: 3,024, Level: 33
    Level completed: 83%, Points required for next Level: 26
    SmoothJohn's Avatar
    Location
    Edmonton, Canada
    Posts
    165
    Thanks
    11
    Thanked 12 Times in 10 Posts

    Re: [PoSt] Philsophy of Statistics

    Just out of curiosity, what do you guys mean by "positivism"? I see it used as a whipping boy in lot's of social science discourse, but it is rarely made explicit enough for decent analysis.

    John

  3. #18
    R must die
    Points: 24,995, Level: 95
    Level completed: 65%, Points required for next Level: 355
    noetsi's Avatar
    Posts
    4,535
    Thanks
    278
    Thanked 730 Times in 700 Posts

    Re: [PoSt] Philsophy of Statistics

    Positivism was the dominant intellectual paradigm in most research areas until the late sixties (and arguably remains so in practice although rejected commonly in theory). Essentially it asserts that reality is defined by what is discoverable through scientific method. If knowledge is not discovered through scientific method than (depending on the author) it either does not exist or is not valid knowledge. Inherent in this, although not part of the philosophy per se, is what is called scientific method - essentially ways we conduct analysis. It includes rules such as letting the data not opinion decide what is correct, avoiding "subjective" analysis, the use of null hypothesis (which have philisophical not just practical elements to it) and so on.

    The post-positivist critique is complex and varied. One element of it argues that it is impossible to avoid subjectivity in analysis, because human beings are inherently subjective. Similarly, it argues that it is impossible to let the data answer questions (critical to positivism's focus on objectivity) because even basic steps such as which data you gather and what questions you ask are subjective. Qualitative analysts argued that many of the rules such as random sampling were invalid for both methodological and philisophical reasons. In general this critique notes that largely hidden in positivism are central assumptions such as support for the existing state of knowledge that are not without challenge - certainly that were judgements as compared to being certain.

    I barely touched on this topic, its incredibly complex. It reached its heights I think from the sixties through the eighties. Today I would guess that most researchers who pay attention to such are post-positivist in theory although commonly positivist in the way they do their research. Beyond the issues I have raised, their is the added related issue of the status in society of researchers and research.
    This was not what we did in logistic regression. Rather, we transformed the conditional expected value, and made that a linear function of X. This seems odd, because it is odd..

  4. #19
    Points: 3,024, Level: 33
    Level completed: 83%, Points required for next Level: 26
    SmoothJohn's Avatar
    Location
    Edmonton, Canada
    Posts
    165
    Thanks
    11
    Thanked 12 Times in 10 Posts

    Re: [PoSt] Philsophy of Statistics

    Thanks noetsi. I have serious doubts that positivism was ever expressed so baldly or believed so naively by any philosopher. It seems at best a caricature of epistemological realism.

  5. #20
    R must die
    Points: 24,995, Level: 95
    Level completed: 65%, Points required for next Level: 355
    noetsi's Avatar
    Posts
    4,535
    Thanks
    278
    Thanked 730 Times in 700 Posts

    Re: [PoSt] Philsophy of Statistics

    I was not trying to claim that it was believed quite that way (although my comments do reflect the views of many of its critics). Just sumarizing some of the central elements in a very general way as expressed by academics - few of whoom were philosophers. I lack the training in honesty in philosophy to express it precisely. Logical positivist, people like Herman Simon took truly extreme positions, most I am sure did not. You might look at (on opposite extremes) the statements of Herman Simon and (as a critic) Lincoln and Guba.

    As is often the case in academics there were pratical elements at play here. Positivism had many useful aspects to the social sciences which felt itself slighted in the public mind (and in funding) relative to the hard sciences. It (many academics believed) increased the legitimacy of social science research if they were conducting "real" scientific research. And to do so, it was felt, they had to use scientific methods (notably statistical analysis) and apply what social scientists felt were the values and behaviors of hard scientists. Whether the hard sciences actually used those methods as rigerously as was perceived in the social sciences is questonable, but it was felt they were in the heyday of positivism which ran roughly from the late forties to mid sixties. This was not concidently an era with extreme confidence in the use of scientific method to address social issues and the beginning of the revolution in computers which made the analysis possible.

    It has been argued, to me with some validity, that it is inherently in academics advantage to support positivism. If there is no objective way to do research, then researchers lack the ability to generate answers that are "better" (and thus more legitimate) in society relative to those that lack the specialized academic training. This of course reduces the validity of research, and particularly social science research, in the general public. By claiming they had a unique, more effective way to generate knowledge researchers were ablet to increase the possibility their research would be accepted and decrease the validity of non-academic commentators. Essentially they were claiming through positivism that they had the only legitimate way to discover reality. Then to empirical research is very hard to justify generally if you don't accept positivist tenets.

    So even those who reject positivism, which is likely most academics today, in practice conduct research as if scientific method was correct.

    You might look at the history of "behaviorialism" the form of positivism that applied to political science and related fields. It reflects well the difference between politics, practical issues, and philosophy in the academic community. A community that is loath to admit it conducts research in fundamentally subjective ways...

    If you think about the assumptions on this board, it is pretty easy to see just how deeply positivism is held to even today. Largely of course without question. Do most here believe that statistical method is inherently objective, that accepting the status quo in theory until disproven is the proper way to do analysis, that statisticans understand reality better than non-statisticans, that rarefied objective facts exist seperate from the meaning attached to them in society? Without really questioning any of that? Probably.

    I know I do, and I am not even a statistician.
    This was not what we did in logistic regression. Rather, we transformed the conditional expected value, and made that a linear function of X. This seems odd, because it is odd..

  6. #21
    Probably A Mammal
    Points: 18,957, Level: 87
    Level completed: 22%, Points required for next Level: 393
    bryangoodrich's Avatar
    Location
    Sacramento, California, United States
    Posts
    2,201
    Thanks
    292
    Thanked 494 Times in 449 Posts

    Re: [PoSt] Philsophy of Statistics

    I think the way Noesti explains positivism borders on scientism. A prominent feature of positivism is merely that knowledge can only be obtained through science as the sole source of authentic information due to its ability to be positively verified. This begs a ton of questions about verification in science. Nonetheless, it was an important part of the social scientific development because it looked the very trite notion that "the natural sciences are successful and can positively verify their results through scientific method; social science can use scientific method to verify their results." Notice how this begs a huge questions about how scientific theories can be verified by the instance of any experiment, and it assumes a continuum between physical and social phenomena in a very big way. Extending this, but certainly within its own lineage, came the logical positivists that sought to stay away from the sort of ontological commitments that make positivism seem so much like scientism in the way Noesti expounded--i.e., that because the only authentic knowledge is scientific, science is the only source of reality. This implies that ontology is founded in epistemology, which it need not be. Logical positivists instead focus on logic as the foundation of scientific reasoning and inference. They accept the positivist epistemology, make no ontological commitments, but only seek to find logical foundations. The key in the logical positivist philosophy is not verification per se, but logical certainty in our assertions. In fact, this is a bolder claim than the positivists could make! It is widely accepted that Karl Popper did a good job dislodging PoS from this view by recognizing the importance of falsification. In particular, this led PoS to look at "what is science?" and try to find what merits are required for an activity to be considered "scientific." Positive verification can never lead one to justify their theory. It is entirely question begging or moot. Of course, the extreme of Popper's position is that we cannot make any verification, only falsification, and all that is necessary is falsification. Kuhn, Lakatos, Lauden, and Feyerabend, among others, critiqued this over the decades revealing that our view of what is "science" is quite diverse and how one goes from evidence to theory and back again is not as objective and rigorous as the logical positivists would have us believe: there is no inductive logic. These postpositivist critiques are not uniform because it's like trying to say atheism as an anti-doctrine of religion is an ideology. It isn't, it's just a rejection of one. Postmodern critiques, however, do have certain features that are important to notice as they are a result of specifically critiquing the positivist and early responses. They highlight that facts don't speak for themselves and they must be interpreted. This requires semantic assumptions about our ways in which we analyze information. Method does not happen in a vacuum. It comes from a semantic structure with a provided syntax that defines what is grammatically correct and what each piece of the vocabulary means. Not every fact, therefore, is objective, because it is biased by the default position one takes regarding the facts, the assumptions behind their models of the world and theory, what other information they have been exposed to, and the community of science at that time. Kuhn especially focused on the cultural aspects of science, and that has played out as a large part of the postmodernist critique, as many people who use it as a pejorative usually do so in the sense that they're implying you think "anything goes because it's whatever people collectively say is true." That's obviously a strawman and ignores the significance of the critique. But now that I've ran along this train of thought for this long, I'll leave you all to

    tl;dr

  7. #22
    R must die
    Points: 24,995, Level: 95
    Level completed: 65%, Points required for next Level: 355
    noetsi's Avatar
    Posts
    4,535
    Thanks
    278
    Thanked 730 Times in 700 Posts

    Re: [PoSt] Philsophy of Statistics

    I was actually talking more about how philosophy has actually been applied historically in the social sciences than the philosophy itself (which I am much less aware of). What positivism is in a purely intelectual sense is very different than what the term has meant in the social sciences - that is what academics in those areas meant when they spared over positivism. This may well be one of those concepts (like culture) that changed its nature fundamentally when it was grafted from one area (in this case philosophy) to method (in the social sciences).
    This was not what we did in logistic regression. Rather, we transformed the conditional expected value, and made that a linear function of X. This seems odd, because it is odd..

  8. #23
    Probably A Mammal
    Points: 18,957, Level: 87
    Level completed: 22%, Points required for next Level: 393
    bryangoodrich's Avatar
    Location
    Sacramento, California, United States
    Posts
    2,201
    Thanks
    292
    Thanked 494 Times in 449 Posts

    Re: [PoSt] Philsophy of Statistics

    That is true noetsi. My friend has been more involved in the social sciences than I have (getting her master degrees in economics and political science--international relations, specifically). From reading some of her books and my own studies in the social science, at least in the textbook perspective, they do tend to make a point about positivism. Historically it did play a significant role in the way social scientists perceived their field. This is quite different from positivism in PoS as a study of philosophy. This is true of all PoS, though. The work Sober does in the philosophy of biology, for instance, doesn't necessarily reflect any great trends in actual biological work or the culture thereof. Similarly, the work of Cartwright hasn't particularly driven econometrics anywhere nor has any work of philosophers in physics driven the way we perceive ontological claims brought out by the latest works at CERN. These are all facets of actual scientific work performed by scientists. Though, to be fair, that line has blurred at times and especially by the training of the scientists and the extent of theoretical work they have done (e.g., Friedman and Keynes made considerable contributions to what we would now call the philosophy of economics while actually being working economists, but let us not forget economics is a very young field to which everybody making serious contributions were at the fore!).

    I think this example should be considered significantly given that if we are discussing the philosophy of statistics, we have to be clear are we talking about it as a philosophy or merely speculating on the theorizing of practicing statisticians? They both have merits, and we're certainly all more involved with the latter, but the former is most aptly what I would think this thread is about.

  9. #24
    R must die
    Points: 24,995, Level: 95
    Level completed: 65%, Points required for next Level: 355
    noetsi's Avatar
    Posts
    4,535
    Thanks
    278
    Thanked 730 Times in 700 Posts

    Re: [PoSt] Philsophy of Statistics

    Philosophy tends to be its own distinct world in academics. A central feature of positivism as applied to the social sciences after the second world war was to move away from philosophy to empirical analysis (which positivist in social science suggested was possible - that is basing answers purely on empirical investigation was possible and desirable). This is one of the things that gets hammered by post(anti) positivist of course and with good reason. It is impossible to conduct empirical research apart from theory and it would make little sense to do so.

    Again I think it is critical to understand that positivism as applied to social sciences had much to do with politics, those who drove it believed that academics would have more respect and funding if it was perceived as a science (by which they meant an emphasis on empirical data and scientific method). And, in the context of the forties and fifties that was a reasonable view. I don't think that philosophy per se had much to do with this aside from a view like Herbert Simon and his impact had little to do with his logical positivism which few were aware of.

    My guess is that most academic statisticians pay little attention to the underlying issues we are talking about. Because scientific method and empiricalism is so well accepted as to be invisible in their world (as in the hard sciences generally). It is only in areas like the social sciences where it was new and commonly not accepted that positivism became a major issue.
    This was not what we did in logistic regression. Rather, we transformed the conditional expected value, and made that a linear function of X. This seems odd, because it is odd..

  10. #25
    Human
    Points: 7,379, Level: 57
    Level completed: 15%, Points required for next Level: 171
    GretaGarbo's Avatar
    Posts
    862
    Thanks
    286
    Thanked 299 Times in 268 Posts

    Re: [PoSt] Philsophy of Statistics

    When I first saw this thread I thought that it would be about the difference in statistical “philosophy” between Fisher, Neyman-Pearsson and Baysians. So I looked at a paper by Lemann that you can find here .

    The Lemann paper is interesting in itself but there was also mentioned a few books about “real” philosophy. I am just trying to read the latest one, but I can't really say if the book is good or bad. I don't know how much I understand either. Probably less then I think. Now, it is more true than ever that “all I know is that I know nothing” (quoted from a famous person).

    I just want to share the refs for these books:

    Seidenfeld, T. (1979) Philosophical problems of Statistical Inference, Boston

    Kyburg H. E. Jr. (1974) The Logical Foundations of of Statistical Inference, Boston

    Hacking, I. (1965) Logic of Statistical Inference, New York

    Braithwaite, R. B. (1979) Scientific Explanation, Cambridge

    -----

    Lemann, E. L. (1993) The Fisher, Neyman-Pearsson Theories of Testing Hypothesis: One Theory of Two, JASA

  11. The Following 4 Users Say Thank You to GretaGarbo For This Useful Post:

    bryangoodrich (10-12-2012), SmoothJohn (12-19-2012), trinker (10-12-2012), vinux (12-18-2012)

  12. #26
    Dark Knight
    Points: 2,450, Level: 29
    Level completed: 50%, Points required for next Level: 150
    vinux's Avatar
    Posts
    1,896
    Thanks
    45
    Thanked 193 Times in 160 Posts

    Re: [PoSt] Philsophy of Statistics

    Thanks Greta for the references.

    I have recently read novel by author of Dilbert, God's debris (Reading the sequel now). Easy read and nice one (trinker may hate, Jake may like). Law of probability is the main theme of the novel. Not dealing with any statistical concepts except using the name "probability".
    In the long run, we're all dead.

  13. #27
    Dark Knight
    Points: 2,450, Level: 29
    Level completed: 50%, Points required for next Level: 150
    vinux's Avatar
    Posts
    1,896
    Thanks
    45
    Thanked 193 Times in 160 Posts

    Re: [PoSt] Philsophy of Statistics

    After reading some PoS article, now I am able to follow this thread. Now I understand the jargon like epistemology, ontology, positivist, naturalist,realism, holism,.... It was very difficult to grasp in the beginning because nature of the topic. I realize that I need to unlearn some of the high school science. The debatable questions like "what is science ?" "pseudo science". "social science", and problems in categorizations were too confusing initially. Best part is reading again and again (same article) gave me lot of insights. Currently I am reading book by Rosenberg, Philosophy of social science

    Now come back to statistics. There is a wiki page on philosophy of statistics we could also add philosophy of probability in our discussions. I found the references are useful (some of them were already mentioned in the previous posts). I have started reading the articles one by one. I will try to answer some of the smoothjohn's questions.

    There is a philosophy forum in stackexchange. Philosophy of mathematics is discussed there. As of now I haven't seen PoSt related questions. I hope there would be many of such questions in the future.

    I found some of the hacking lectures in youtube. Here is one and there are more.
    In the long run, we're all dead.

  14. The Following User Says Thank You to vinux For This Useful Post:

    GretaGarbo (01-07-2013)

  15. #28
    Dark Knight
    Points: 2,450, Level: 29
    Level completed: 50%, Points required for next Level: 150
    vinux's Avatar
    Posts
    1,896
    Thanks
    45
    Thanked 193 Times in 160 Posts

    Re: [PoSt] Philsophy of Statistics

    Very Useful information on intersection of philosophy and statistics.

    1. Philosophy of Statistics (Course offered by Deborah G Mayo)
    2. Philosophy and the practice of Bayesian statistics (with all the discussions!) (Andrew Gelman)
    In the long run, we're all dead.

  16. The Following User Says Thank You to vinux For This Useful Post:

    bryangoodrich (02-09-2013)

  17. #29
    Probably A Mammal
    Points: 18,957, Level: 87
    Level completed: 22%, Points required for next Level: 393
    bryangoodrich's Avatar
    Location
    Sacramento, California, United States
    Posts
    2,201
    Thanks
    292
    Thanked 494 Times in 449 Posts

    Re: [PoSt] Philsophy of Statistics

    I don't know if I mentioned it before, but to me I think philosophy of statistics has a solid place in the epistemology studied in the various philosophy of sciences, because (1) statistics is used primarily in the various sciences, and (2) statistics is principally concerned with obtaining information ("knowledge"), usually through inference. That's simplistic, I know, but my point is that when exploring philosophy, be cognizant of things not about statistics directly, but which are still relevant. I've mentioned Elliot Sober's book before. It's basically a book about epistemology in the area of biology (philosophy of biology). The entire first chapter is about statistics, though. I've been reading a bit of philosophy of economics stuff, and I think it is also relevant since there is a lot of talk about causality (can economics achieve it?), and causality in statistics or inductive inference is an important topic. That raises another point, a lot of the time they don't even talk about statistics when they are talking about statistics. Usually they generalize to induction, inductive inference, or inductive logic when talking about the same thing, essentially. One thing I find interesting is the role statistics plays in theory development. By theory in philosophical terms, I mean "a collection of statements that explain a body of disparate phenomena." Moreover, we might append to this definition that theories are truth-seeking. By this I mean that we don't simply take those statements to be true or serve a pragmatic (see "predictive") purpose, but that those statements are true about the world (phenomena) that they explain. This is a point of consternation in economics because often the theory is making false statements, particularly when they're behavioral--e.g., people do not behave rationally in the way economics demands for their theory to be true about markets.

    Anyway, I could lecture for days about philosophy of economics Thinking about it now, it's kind of disgusting how many books on it I've read (especially if I include PoS generally) lol

    My current reading material: http://www.amazon.com/Preference-Cho.../dp/1107695120
    You should definitely use jQuery. It's really great and does all things.

  18. #30
    Dark Knight
    Points: 2,450, Level: 29
    Level completed: 50%, Points required for next Level: 150
    vinux's Avatar
    Posts
    1,896
    Thanks
    45
    Thanked 193 Times in 160 Posts

    Re: [PoSt] Philsophy of Statistics


    I have recently gone through some of the articles on genrel laws and causality. Until the "smoking debate" attitude of statistics was not used to establish causality. Fisher or statisticians in the beginning of 20th century considered "statistics is only about association". Last 3-4 decades, different causality establishing techniques developed in statistics. Rubin causal model, structural equation models, granger causality are few of them. counter-factual definition and/or time lag analysis are the key elements to establish the causality. It is not statistics alone is used to establish the causality. Statistics and the context mechanism (experience based and mechanistic) is used to establish the relationship.

    I yet to read PoSt articles. Still spending time on PoS articles. Planning to read some articles related to my research (rationality; fall under philosophy of economics).

    I found following paragraph (by FA Hayek) interesting.
    How little statistics can contribute, however, even in such cases, to the explanation of complex phenomena is clearly seen if we imagine that computers were natural objects which found in sufficiently large numbers and whose behavior we wanted to predict. It is clear that we should never succeed in this unless we possessed the mathematical knowledge built into computers, that is, unless we knew the theory of determining their structure. No amount of statistical information on the correlation between input and output would get us any nearer our aim. Yet the efforts which are made on a large scale with regard to the much more complex structures which we call organisms are of the same kind. The belief that it must be possible in this manner to discover by observation regularities in the relations between input and output without the possession of an appropriate theory in this case appears even more futile and naive than it would be in the case of the computers.
    It is from a 1978 article named "The theory of complex phenomena"
    In the long run, we're all dead.

+ Reply to Thread
Page 2 of 3 FirstFirst 1 2 3 LastLast

           




Tags for this Thread

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts






Advertise on Talk Stats