Re: help on factor analysis/non-normality
to amplifiy a bit, the interpretability of regression tends to go down as the assumptions of normality and homogeneous variance are markedly different from reality. You can still go through the calcualtions but the interpretation of results gets tricky. Factor analysis is a sort of regression analysis and so suffers in the same way from break downs of assumptions. Rich Ulrich wrote: > On 1 Mar 2002 04:51:42 -0800, [EMAIL PROTECTED] (Mobile Survey) > wrote: > > > What do i do if I need to run a factor analysis and have non-normal > > distribution for some of the items (indicators)? Does Principal > > component analysis require the normality assumption. > > There is no problem of non-normality, except that it *implies* > that decomposition *might* not give simple structures. > Complications are more likely when covariances are high. > > What did you read, that you are trying to respond to? > > > Can I use GLS to > > extract the factors and get over the problem of non-normality. Please > > do give references if you are replying. > > Thanks. > > -- > Rich Ulrich, [EMAIL PROTECTED] > http://www.pitt.edu/~wpilib/index.html = Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at http://jse.stat.ncsu.edu/ =
Re: EDA
Data mining , by and large, seems to use fairly conventional multivatiate stats tools along with a bunch of clustering procedures. In addtion there is a lot of use of neural nets (mostly as a lazy man's tool or a last resort, but occasionally sensibly). Data prep. (including transformations) seem to be a necessity. A good starter book is "Data Preparation for Data Mining" by Dorian Pyle. It is equivalent to the first part of a low level intro stats book and is mainly concerned with assessing the distributions, variance structure, etc. before deciding to press ahead. I have not so far seen a sensible book on data mining itself. Definitely none equivalent to the many fine texts out there on ultivariate statistics. Many of the DM books are sales blurbs for one or another black-box package. things should change for the better in a couple of years. SR Millis wrote: > I'm looking for recommendations for recent books and papers on basic > techniques for exploratory data analysis. > > Thanks, > SR Millis > > = > Instructions for joining and leaving this list and remarks about > the problem of INAPPROPRIATE MESSAGES are available at > http://jse.stat.ncsu.edu/ > = = Instructions for joining and leaving this list and remarks about the problem of INAPPROPRIATE MESSAGES are available at http://jse.stat.ncsu.edu/ =
Re: Factor analysis - which package is best for Windows?
you may wish to consider NCSS (they have a web site) provides essentially the same output as SAS but is run from templates not SAS language. Less expensive, good documentation, excellant support. However does not provide an audit trail--a necessary feature for some governmental / legal groups. PeterOut wrote: > [EMAIL PROTECTED] (Magill, Brett) wrote in message >news:<[EMAIL PROTECTED]>... > > Also check out R, a GNU implementation of the S language, most prominently > > known through its use in S-Plus. R is a fully featured statisitical > > programming environment. In its MVA (Multivariate) package, it includes > > routines for factor analysis using maximum liklihood estimation with varimax > > and promax rotations. > > > > I have installed R1.3.0 on my Windows system and have noted that MVA > is an add-on. The FAQ tells how to obtain these add-ons but only for > UNIX. Is this add-on actually available for Windows? If so, how do I > obtain it? > > Thanks, > Peter = Instructions for joining and leaving this list and remarks about the problem of INAPPROPRIATE MESSAGES are available at http://jse.stat.ncsu.edu/ =
Re: Methodology development: Open vs. Proprietary
No procedure is bullet-proof--if I don't know something about the "engine under the hood" I can't know when it is likely to fail or provide stupid results. On real world problems where there is some debate concerning validity of results, "secret" proprietary output is just meat for sharks like me. "T.S. Lim" wrote: > In the Data Mining world that is dominated by Computer Scientists, the > methodology behind the software packages sold/licensed in the market > is often proprietary. Take, for example, the classification and > regression trees software package CART(r). The basic idea behind > CART(r) is the algorithm proposed by Breiman, Friedman, Olshen, and > Stone (1984). However, there has been quite a few proprietary > improvement in CART(r) so that you can no longer know for sure what's > going on inside the software package. The same is true for C5.0/See5 > (another classification trees software) that supersedes C4.5. > > When dealing with proprietary methodology, it's (practically) > impossible to study the properties of the method > thoroughly. Personally, I feel uncomfortable using a method that can't > be evaluated objectively by fellow researchers. It may be OK if the > application has nothing to do with human experimentation (as in > Biostatistics). Since most (if not all) applications of Data Mining > are in commerce, the risk of using unproven methodology that hasn't > been extensively scrutinized may be acceptable. > > Perhaps this joke is true after all: when a Statistician gets an idea, > she/he'll write and publish a paper while when a Computer Scientist > gets an idea, she/he'll form a company. :) > > Comments? > > -- > T.S. Lim > [EMAIL PROTECTED] > www.Recursive-Partitioning.com > > > Get paid to write review! http://recursive-partitioning.epinions.com = Instructions for joining and leaving this list and remarks about the problem of INAPPROPRIATE MESSAGES are available at http://jse.stat.ncsu.edu/ =
Re: stat packages
NCSS is beautiful in its ease of use and robustness. It comes with a set of printed documentation that is really a series of applied stats books including literature references. I have used the current version for about a year. It has a few minor flaws in that it occasionally locks up or boots you out--I don't know if it is a MS prob or an NCSS prob. But i strongly recommend it in terms of easeof use, comprehensiveness, high precision and telephone support by the guy who put it together, Jerry Hintze. Charlie wrote: > I was a SAS user from the mainframe days until about 1995. I liked the > software but the yearly lease was prohibitive. > > I tested several stat packages at that time and settled on Statistica. > I have no regrets. Their ads are every bit the truth. It's a great > package, well worth the $1,000. Rumor has it the package is alien > technology from the Roswell, NM crash. This is probably not true > because the statistics are very up to date. > > I have a colleague who swears by NCSS, but I've never used it. The > $300 price tag sounds cheap for what it's supposed to do. Maybe it's > alien technology. > > I've never warmed up to Minitab. I guess I know too many rabid Penn > Staters who think it's the second coming of SAS. > > Statgraphics? Well I've never recovered from our first meeting in the > late 1980s. > > Charlie > > On Fri, 08 Sep 2000 19:53:04 GMT, "AJ" <[EMAIL PROTECTED]> > wrote: > > >I am interested in opinions on the Statistica package. I have always used > >SPSS, but now that I need to buy my own program, I am intrigued by > >Statistica. Not surprisingly, their ads are very compelling. I need a > >general, broad-based package with basic stats, GLM, regression, survival > >analysis, and graphics. I have used SPSS since the mainframe days, but I am > >Statistica (and Systat) appear to provide excellent value. I am a > >behavioral science researcher with a moderate to strong background in > >applied multivariate analysis (not a statistician). Any comments? > >Thanks. -- Al J. > > > > > > > > Charlie Kufs > ~ > TerraByte Corporation > 137 Summit Avenue > Willow Grove PA 19090-3108 > [EMAIL PROTECTED] > http://terrabyte.ws > ~~ = Instructions for joining and leaving this list and remarks about the problem of INAPPROPRIATE MESSAGES are available at http://jse.stat.ncsu.edu/ =
Re: John Tukey
a great spirit. An ornament to the Profession. A person who made all of our lives easier. A person who wrote with the gusto and spirit of an enthusiast. A Hero. Robin Becker wrote: > In article <[EMAIL PROTECTED]>, Petr Kuzmic > <[EMAIL PROTECTED]> writes > > > > > >Donald Macnaughton wrote: > >> John Wilder Tukey died last night > > > ... > very sad news > -- > Robin Becker = Instructions for joining and leaving this list and remarks about the problem of INAPPROPRIATE MESSAGES are available at http://jse.stat.ncsu.edu/ =
Re: an open review process in Statistics journals?
The peer review process is imperfect however you have to take a stoic approach and be patient. I am nervous when i get totaly positive reviews because it may mean that the reviewers are not doing their job or that I have submitted a paper to a "hungry journal". I would say that, in my field, about 1/3 of the reviewer simply do not get the poin and read into the paper things that are not there. I see that as a failure on my part in clarity and so revise wiht that in mind. Some times reviewers criticism is based on outright error in fundamentals, in which case a letter to the editor, calmly pointing this out is appropriate. In a very few instances a bad review will come from someone defending a bit of intellectual capital come what may. again a well thought out letter to the editor is one possibility. I have had a few papers that required "major revision" or even complete rewriting. By and large the papers are the better for it. So hostile reviews, although irritating and sometimes off the mark, can result in a clearer paper. Please do not stop writing and submitting papers nor make fundamental changes just to please the reviewers. Xie Min wrote: > > T.-S. Lim ([EMAIL PROTECTED]) wrote: > : I'd like to hear others' opinions regarding making the review process for > : submitting papers to journals totally open. In my very limited experiences, > : I've encountered referees who don't know what they're talking about. They even > > I think this is not uncommon. On the other hand, a journal also depends on > the chief editor (who selects the referees, and also judges if a referee > report is good or not) or Associate Editor (in some cases, the editor > passes to the associate editor who is more knowledgeable with the topic, > and the associate editor selects referees). > > : make silly comments. IMHO, they would think twice before writing any silly > : comment if they know that their names would be made known to the author(s). > > The problem is you will then get positive reports with a long review > process... I think most people will then decline to be a referee if > they are not happy with the paper. > > : I'd like to venture that an open-review process would increase the overall > : quality of the journal. Thank you for reading this. Email me if you'd like to > : have a private discussion. > > If you like open-review, then an open discussion should be better... > > Anyway, either is acceptable (if handled properly), but an open > review will probably cause more problems, unhappiness and conflict > (we are all human being...). You can imagine what will happen if > the presidential election is totally open (we know whom other > voted...).
