I was amused to learn, for example, that many august statistical techniques like analysis of variance were created so that someone could figure out how much artifical cow poop to spread over an acre of farm land. The beginning of the story differs from one account to another. The Lady Tasting Tea is not a book of dry facts and figures, but the history of great individuals who dared to look at the world in a new way. Mostly useful for the biographies of these historical figures. Needless to say, there was a dissonance between Salsburg's excitement and my dull incomprehension of what was so exciting. He knew that how the data was gathered and applied was as important as the data themselves.
Though not quite on the level of say Bill Bryson's A Short History of Nearly Everything, this book does a decent job of layering those pedestrial and alltogether human eccentricities over the enormity of the scientific accomplishments they created. The crude suicide rates data from the period of 1991-2013, which recruited nine urban and 14 rural areas in Taiwan, were extracted from the Taiwanese national mortality data file. In Salsburg's defense, perhaps there just weren't as many colorful personalities in statistics as in physics or mathematics; one gets the impression of a rather dour bunch on the whole. For a new edition, Salsburg might want to do something like that. As someone who wishes I were a real mathematician, it's a little depressing for me to read about a genius like Kolmogorov who came up with interesting results in every field he touched ever since childhood.
This is not, however, a book of mathematical theory. I found it an enjoyable book to read. Collecting all the works and achievements of vastly different authors of many backgrounds and summarize their contribution to the Statistical Revolution in a cohesive narrative? I learnt a lot as well. High divorce and unemployment rates resulted in increased suicide rates in men in the city, whereas emotional distress was the main cause of suicides in men in rural areas. From a conceptual perspective, randomization tests are based on random assignment and permutation tests are based on random sampling. You first mark four dots to make the corners of a box.
Our reliance on group-averaged data creates a dilemma. Particularly well told is the story of Ronald Fisher, the double genius who founded both mathematical statistics and mathematical genetics. . The author also paints a great picture of how rich and varied the field of statistics is, and what interesting people have contributed to it. But it took Fisher to make clear the distinction between a parameter's true value, and your estimator for that parameter the formula you feed data into , and your estimate of the parameter the actual value you got using a particular dataset.
We review why degrading p-values into 'significant' and 'nonsignificant' contributes to making studies irreproducible, or to making them seem irreproducible. The author also brings up the philosophical matters that continue to bedevil statistics, such as the meaning of a probability. But in practice, measurements lacked precision. It's not like there aren't plenty of those out there. The book is organized according to topics in statistics including biographical sketches of the people important to the development and application of each of the topics. In particular, such solution is possible if the data distribution obtains a so-called conjugate prior.
For example, a brilliant agricultural statistician named Chester Bliss couldn't find a job in America during the Depression, so Fisher landed him a post at the Leningrad Plant Institute. Too often we learn concepts and methods that are popular today without understanding why we use them or how they developed. This is inconsistent with the conclusion. To encounter Bayes's theorem or any number of other statistical ideas and see not a single formula or mathematical expression was to me like reading a joke book without any jokes in it. So, the communists left him alone for months until they eventually realized that while he wasn't a spy, he was an anti-communist. This is what happens when one tries to apply statistical models to real-life problems.
I highly recommend this book for all who teach stats as well as for all who have an interest in stats. He also covers other techniques than the ones I mentioned above, such as experimental design, non-parametric statistics, computer intensive methods such as the bootstrap, and Bayesian statistics. For Fisher had brought to the field of statistics an emphasis on controlling the methods for obtaining data and the importance of interpretation. The book at its best in the beginning, before branching off in every which way, when examining the very beginning of the field, with Gosset and Pearson and Fisher. Fisher later discussed the benefits of more trials and repeated tests. Classification models with a decision threshold different from 0.
Salsburg is very careful to not cross the line between giving details and being too technical, so the concepts and ideas are shown in great detail but without mathematics, what makes possible a fluid and smooth reading. What is the probability that such a result comes to us by chance rather than by causation? I was fascinated by the newness in this field. We conclude that whatever method of statistical inference we use, dichotomous threshold thinking must give way to non-automated informed judgment. For Fisher had brought to the field of statistics an emphasis on controlling the methods for obtaining data and the importance of interpretation. In order to appreciate the way it could suit laboratory medicine, it is necessary to understand the philosophy behind it, and in turn how it stemmed and differentiated along the history of classical hypothesis testing. In chapter twenty-four, we learn how Japanese industry became known for superior products notably automobiles by heeding the advice of W. It makes statistics a lot more interesting than just reading equations.
Even with three years of graduate statistics from a social science perspective , I often found myself unsure of his explanations. These discussions help to expose both errors and new ideas. I'll be able to share more interesting information when teaching stats in the future. As a result, the posterior distribution of the parameter of interest will represent the elicited prior distribution that does not assume any particular parametric form. I would love to find recordings but googling does not help, although some transcripts might be in if I can find a library with access to this in its database.