livres. lus. approuvés.
Bienvenue chez Bibliomania, le spécialiste en ligne du livre de seconde main
FR  •  NL
Panier
0
Statistical methods in experimental physics
Cartonné / 296 pages / édition de 1977
langue(s) : anglais
ISBN : 0720402395
EAN : 9780720402391
dimensions : 265 (h) x 198 (l) x 20 (ép) mm
poids : 840 grammes
Cet ouvrage n'est
pas disponible
actuellement sur
Bibliomania

commandez-le sur

à partir de EUR 45,58

This course in statistics, written by one statistician (K.T.E.) and four high-energy physicists, addresses itself to physicists (and experimenters in related sciences) in their task of extracting information from experimental data. Physicists often lack elementary knowledge of statistics, yet find themselves with problems requiring advanced methods - if adequate methods at all exist. To meet their needs, a sufficient course would have to be very long. Sudi courses do indeed exist [e.g. Kendall], only the physicists usually do not take the time to read them.

We attempt to give a course which is reasonably short, and yet sufficient for experimental physics. This obviously requires a compromise between theoretical rigour and amount of useful methods described.

Thus we are obliged to state many results without any rigorous proof (or with no proof at all); still we have the ambition to present more than just a cook-book of prescriptions and formulae. We omit the mention of many techniques which, in our judgement, seem to be of lesser importance to experimental physics.

On the other hand, we do introduce many theoretical concepts which may not seem immediately useful to the experimenter. This we think is necessary for two reasons. Firstly, the experimenter may need to know some theory or some "generalized methods" in order to design his own methods, experimental physics posing always novel questions. This is a justification for the stress on Information theory (Chapter 5), and for the attempt in Chapter 7 to define a "general" method of estimation. We hope that although the method the reader will arrive at may not be optimal, still it will be useful.

Secondly, the experimenter should be aware of the assumptions underlying a method, whether it be a standard method or his own. It is for this reason that we insist so much on the Central Limit Theorem, which is at the foundation of all "asymptotic" statistics (Chapters 3, 7).
Quoting theorems, we also try to state their range of application, to avoid too careless use of some methods.

Among the underlying assumptions, especially important are the ones about the parent distributions of the data, since they will condition the results. In Chapter 4 we give a catalogue of useful ideal distributions; in real life they may have to be truncated (Section 4.3), experimental resolution may have to be folded in (Section 4.3), detection efficiency may have to be taken into account (Section 8.S). Moreover, the true distribution may not be known, in which case one is led to empirical distributions (Section 4.3), robust estimation (Section 8.7), and distribution-free tests (Chapter 11).

A very common tacit assumption in the everyday use of statistics is that the set of data is large enough for asymptotic conditions to apply. We try to distinguish clearly between asymptotic properties (usually simple whenever they are known) and finite sample properties (which are usually unknown). We also often give asymptotic expansions, in order to indicate how rapidly the asymptotic properties become true.
rechercher des articles similaires par thème: