<- file stat .html -> ******************* Less common areas for .stat. **********

Qualitative data analysis

======================Morris, 30 May 1996==========ssc From: morrisa@netcom.com (morris) Subject: Re: Books for qualitative data analysis Message-ID: <morrisaDs8K6x.L8o@netcom.com> I have found the following references to be very useful. Weller, Susan, & Romney, A. K. (1988). Systematic data collection. Sage qualitative research methods series, number 10. Miles & Huberman (1994?) The title is something like, "Analysing qualitative data," or "Qualitative data analysis." 2nd edition, Sage. It's an excellent book. Bernard, H. Russell. (1994). Research methods in anthropology: Qualitative and quantitative approaches (2nd ed.). Sage. The last is probably the best book on research methods that I've read, and the methods are certainly applicable to any social science research. Bernard discusses topics such as how to take, code, and manage fieldnotes, interviewing, choosing informants, etc., and provides a fresh discussion of other topics more frequently found, such as scales and scaling. I find it to be a very well written book, with clear lucid prose. Morris * * * * * * * * * * * * * * * * * * * * * * * * * * * * *

Control charts

====================William C. Parr, 09 Jun 1996======sse,ssc on SPC From: wparr@utk.edu (William C. Parr) Subject: Re: Wanted: Exercises in Rational Subgroups Message-ID: <wparr-0906962123450001@tchm04a13.rmt.utk.edu> In article <4p6kq8$k9m@usenetw1.news.prodigy.com>, DMTP35A@prodigy.com (Michael Daniels) wrote: > I need some simple exercises to demonstrate the concept of rational > subgroups in control charts. The example should be easy to execute, use > simple materials and clearly demonstrate how different answers to > different questions are dependent on how one subgroups the population. > > - > MICHAEL DANIELS DMTP35A@prodigy.com Michael, you might want to check out the book: The Power of Statistical Thinking, by Mary Leitnaker, Richard Sanders, and Cheryl Hild. It has an excellent chapter (Chapter 6) on the subject (more than I've seen in any other book, by far!). Publisher: Addison Wesley. Next choice: Understanding Statistical Process Control and Advanced Statistical Process Control, by Wheeler & Chambers and Wheeler respectively, both published by SPC Press. I'd be so bold as to say that all three books should be owned by anybody wishing to go deeper on the subject than the usual level of a little philosophy and some mechanics. All three books get beyond the "SPC as a method for rapid detection of assignable causes" and into "SPC as part of a strategy for process study leading to process improvement." Enjoy, Bill Parr * * * * * * * * * * * * * * * * * * * * * * * * * * * * *

Circular statistics. Zar

=====================Jerrold Zar, 03 Mar 1996========ssc Message-ID: <s139ae78.026@WPO.CSO.NIU.EDU> From: Jerrold Zar <T80JHZ1@WPO.CSO.NIU.EDU> Subject: Circular Statistics Books/Reviews -Reply I would say the most commonly cited publications are these: Batschelet, E. 1981. Circular Statistics in Biology. Academic Press. Fisher, N. I. 1993. Statistical Analysis of Circular Data. Cambridge University Press. Mardia, K. V. 1972. Statistics of Directional Data. Academic Press. Mardia, K. V. 1981. Directional statistics in geosciences. Communic. Statist. -- Theor. Meth. A10:1523-1543. Upton, G. J. G. and B. Fingleton. 1989. Spatial Data Analysis by Example. Vol. 2. Categorical and Directional Data. John Wiley & Sons. Chapter 9. Zar, J. H. 1996. Biostatistical Analysis. Prentice Hall. Chapters 25 and 26. * * * * * * * * * * * * * * * * * * * * * * * * * * * * *

Case-matching software

========================Erik Bergstralh, 07 Jun 1996=======ssc From: Erik Bergstralh <bergstra@mayo.edu> Subject: Re: Case Matching Software Message-ID: <31B8A4A3.584B@mayo.edu> Richard Mccleary wrote: > > Does anyone know whether SAS, Stata, or SPSS/PC have routines for > matching a sample? I have a file with 51 cases and 659 controls. I want > to select the 51 (of 659) controls that are closest to the 51 cases on > some variable or variables. We have developed a SAS macro(%match) to do this sort of thing. It incorporates both greedy matching and the optimal alogrithm of Rosenbaum(JASA 1989). It is briefly described in the May 1996 issue of Epidemiology. If anyone would like a copy please contact me by e-mail. Erik Bergstralh, Biostatistics, Mayo Clinic * * * * * * * * * * * * * * * * * * * * * * * * * * * * *

Environmental statistics.

=====================James Ssemakula, 07 Sep 1995========sse Message-ID: <950907171109.2120f04c@ucrac1.ucr.edu> From: JAMES@ucrac1.ucr.edu Subject: RE: Environmental Statistics R.O Gilbert 1987. Stat. Meth. for Env. Pollution Monitoring EPA 1989. Stat. anal. of grnd-h20 monitoring data at RCRA facilities. PB89-151047 (republished by NTIS?) R.M Berthouex 1994. Stats for Env. Engrs. P.K. Sen (ed) 1985 Biostats: stats in biomed, pub helth & env sciences G.P. Patil & C.R. Rao (eds) 1994 Env. stats ----------- 1993. Multiv. Env. stats R. Gothern & n.P Ross (eds) 1994 Env. Stats, assessment & forecasteing W.R. Ott 1995 Env. Stats & data analysis A.T Walden & P. Guttorp (eds) 1992 Stats in Env. & Earth Sciences S.M. Gertz & M.D. London 1984. Stats in the Env. Sciences. ASTM Spec. Tech. Rpt. #845) J.H Hwang 1987. Description of univariate stat. models for use in Env data analysis. NCASI Tech Bull #530 I regret the use abbrv. but hope you can decipher them. james ssemakula uc riverside ps: you want to bone up on the lognormal distribution, censored and/or truncated data analysis * * * * * * * * * * * * * * * * * * * * * * * * * * * * *

Linkage analyses (genetics), Where?

=====================Charles Berry, 29 Feb 1996========ssc From: cberry@tajo.edu (Charles C. Berry) Subject: Re: Linkage analyses, Lod scores, pairs of sibilings, Ott method Message-ID: <4h50tr$afr@news1.ucsd.edu> Jordi Balanyà (scimed@readysoft.es) wrote: : I would like to know about this type of statistical analysis (Linkage : analyses, such as Lod scores, pairs of sibilings or the Ott method. As : far as I know these analyses are useful in molecular genetics to stablish : if a mutation is linked to a particular desease. : Also I would like to know if there is any program for PC that can do it : (currently I'm working with SPSS). Check out: http://hacuna.ucsd.edu/genepi/index.html For PC programs, see http://hacuna.ucsd.edu/genepi/software/os.html click on OS2 and/or DOS Enjoy! -- * * * * * * * * * * * * * * * * * * * * * * * * * * * * *

Nonlinear confidence regions. Chandler.

=====================John Chandler, 10 Apr 1996========ssc,smn-a Newsgroups: sci.math.num-analysis,sci.stat.consult Subject: Re: sum of exponentials Message-ID: <4khg8e$ca@news.cis.okstate.edu> In article <Pine.OSF.3.91.960410085525.14729A-100000-100000@io.uwinnipeg.ca>, Bill Simpson <wsimpson@uwinnipeg.ca> wrote: > >> The computation and reporting of symmetric, linearized parameter >> errors in nonlinear least squares fits is almost universal, >By this do you mean: >SEs calculated by Wald method, SE=sqrt(diagonal(invert(Hessian)))? Roughly, yes. If you don't know the errors in the data, you will have to rescale the parameter errors according to the results of the fit. See the references below. >> but is in many cases horrible statistical practice >> and should never be done unless it has been shown >> numerically that the errors are indeed nearly symmetric. >> >> There are two methods of computing nonsymmetric errors: >> 1) support planes, and >Does this mean: >e.g. when doing maximum likelihood (instead of least squares) find >parameter values that bring the log likelihood down from its value at >maximum to maximum-k, where k depends on confidence level. For 95% CI use >k=1.92 (95th percentile of chi-squared, df=1). Yes, but you must not just move parallel to the parameter axes from the point of least squares (maximum likelihood) to find these points. That will drastically underestimate the errors. You must move a parameter, fix its value, and optimize in that plane. Then iterate until you find a plane perpendicular to the coordinate axis in which the minimum sum of squares (maximum likelihood) has the desired value. That is a support plane for that parameter. >> 2) Monte Carlo simulation. >> One or both of these should always be used in any nonlinear >> regression analysis. The use of either method in this >> context is rare. >I would greatly appreciate any refs on implementation of these esp >support planes. Given that one has an optimization routine that will >find the max log likelihood, how to find those pts bracketting the >maximum but k units lower? You must be able to move, fix a parameter, and optimize under that constraint. With some optimization routines, fixing a parameter is not very convenient, but it is always possible. You may have to squeeze out that parameter, repack the vector of parameters into a vector of length n-1, call the optimizer to solve a problem of size n-1, and, in the subprogram that computes the function and gradient, use the (n-1)-vector plus the fixed parameter value to compute the function. The fixed value, parameter number, etc. can be communicated directly from the main program to the function subprogram using labelled COMMON in FORTRAN, global variables in Pascal, etc. For FORTRAN users, file marq.f in directory /pub/jpc, available free via anonymous ftp at a.cs.okstate.edu, contains MARQ and FIDO. MARQ is a nonlinear least squares routine using the Gauss-Newton, modified Gauss-Newton, or modified Marquardt methods. FIDO fixes parameters, calls MARQ and does the inverse interpolation (rootfinding) necessary to find the support planes. (Eliminate all // and $ job control language and move the data points at the end of the file to a separate file, if necessary.) MARQ is reasonably fast. FIDO has to call MARQ many times, and hence is slow. References on confidence regions and support planes include: "Data Reduction and Error Analysis for the Physical Sciences" by Philip R. Bevington and D. Keith Robinson, McGraw-Hill, 1992 (see p. 212 ff., for example) "Statistical Methods in Experimental Physics" by W. T. Eadie, D. Drijard, F. E. James, M. Roos, B. Sadoulet, North-Holland/American Elsevier, 1971, Chapter 9 "Parameter Estimation in Engineering and Science" by James V. Beck and Kenneth J. Arnold, John Wiley and Sons, 1977, Section 7.7.4 (Section 7.6.4.1 is wrong, incidentally. No correct Marquardt routine will reproduce Figure 7.11b even qualitatively. Marquardt's method is much more robust than is implied by that figure.) "Nonlinear Estimation" by Gavin J. S. Ross, Springer-Verlag, 1990 See also the articles referenced in these books, of course. See also any recent book on nonlinear estimation, nonlinear regression, etc. For example: "Nonlinear Statistical Models" by Ronald A. Gallant, John Wiley and Sons "Nonlinear Regression Analysis and its Applications" by D. M. Bates and D. G. Watts, John Wiley and Sons "Nonlinear Regression" by G. A. F. Seber and C. J. Wild, John Wiley and Sons "Nonlinear Regression Modeling" by D. A. Ratkowsky, Dekker, 1984 -- John Chandler * * * * * * * * * * * * * * * * * * * * * * * * * * * * *

Document by Rich Ulrich. E-mail to wpilib+@pitt.edu

FAQ top.

Ulrich home page.

Ulrich FAQ. http://www.pitt.edu/~wpilib/stats99.html