spence@utcsstat.UUCP (07/09/86)
Gary Perlman has recently compared some aspects of SAS/PC, PowerStat, and his own UNIX|STAT in this news group. UNIX|STAT is unquestionably the cheapest, and SAS/PC is the most expensive. But at current prices can Gary provide support for UNIX|STAT? Will he mail out updates to all users? Will he provide telephone support during business hours? Is his (optional) documentation typeset and bound? How extensive is the documentation? And so on. It's tough to deliver lots of service when it's a hobby and not a business. In my opinion SAS/PC, PowerStat, and UNIX|STAT all have a niche in the marketplace. SAS is a very capable system, but on the PC it is large, slow, and somewhat awkward (not to mention very expensive once you add it all up). And you NEED a hard disk (realistically 20Mb+, if you're planning on doing anything else with your computer). You also NEED 512K, or more, and SAS Institute recommends an 8087. For use in teaching, it is a bit unwieldy in the PC environment. Overall, SAS/PC is a bit of a disappointment on an 8088/86 machine. Maybe when the 386 machines come out ... PowerStat and UNIX|STAT are much more economical in terms of required equipment (and price). They can both run on vanilla PCs with 256K and two floppies. Although PowerStat isn't quite as cheap as UNIX|STAT, it does much more, comes with a 524 page typeset manual, and has the the best user interface of any statistics package on the IBM PC. We've tried PowerStat here at the University of Toronto in several classes: students find it very easy to learn and use. Much more so than SAS (which we formerly used on the mainframe). And I know of people at several other schools (in Canada, the US, and Australia) who have had similar experiences. I'd like to see more good statistical software for PCs--I don't think that we should sit back and let SAS Institute do it all (good as they are). We need innovative approaches like Gary Perlman's UNIX|STAT and Analytical Engineering's PowerStat. Ian Spence {allegra,decvax,seismo,utzoo,watmath}!utcs!utcsstat!spence
perlman@wanginst.UUCP (Gary Perlman) (07/10/86)
Here are some answers to your questions. In article <2282@utcsstat.uucp> spence@utcsstat.UUCP writes: >But at current prices can Gary provide support for UNIX|STAT? I can't provide support, except that I try to help out people if they can't get the package running on a new UNIX machine (there are about 20 hardware-software combinations running |STAT on UNIX). >Will he mail out updates to all users? About once a year, I send out an announcement of what is in the new release, and it is up to the user to order it, if they think the enhancements are worth the extra $15 for MSDOS or $20 for UNIX. People can freely copy the package, both on and off site, so it is not possible for me to get information about updates to all users. >Will he provide telephone support during business hours? This is something that a computer center would want, and something I definately will not provide. >Is his (optional) documentation typeset and bound? It is typeset on paper, and stapled between heavy paper covers. I charge $10 for the handbook, and $5 for printed versions of the online manual entries. To be honest, I don't care if people make copies. >How extensive is the documentation? It is meant for people familiar with running commands like DIR, DEL, COPY, and SORT on MSDOS, or similar commands on UNIX. It can get a lot of whining from undergraduates if they don't get a lot of hand holding, but I have seen many undergraduate courses use it effectively. >It's tough to deliver lots of service when it's a hobby and not a business. You bet, and the person who requests |STAT must be willing to spend the time to read the manuals carefully. However, individual programs do not need much support, because they are UNIX style tools; how much support does someone need for a descriptive stats program that reads all the numbers in its input? One feature I like about the |STAT programs is that they can be used away from data analysis; there are several programs for text manipulation in files. The major problems with learning |STAT is in learning how to make the programs work together, through pipelines. Consider the following command line to make a 5x5 matrix of uniformly random numbers drawn from 1..25 with replacement: probdist random uniform 25 | dm floor(x1*25+1) | maketrix 5 or without replacement: series 1 25 | perm | maketrix 5 These are intuitive to a UNIX user, but it takes a while for students struggling with MSDOS to learn the tricks. >They can both run on vanilla PCs with 256K and two floppies. |STAT runs on vanilla MSDOS, not just IBM PC's. People have told me they have the programs running on AT&T, DEC, Tandy, adn Wang PC's. You don't even need two floppies, but the thought of doing anything on a PC with just one floppy make me queasy. >We've tried PowerStat here >at the University of Toronto in several classes: students find it very easy to >learn and use. Much more so than SAS (which we formerly used on the >mainframe). Although I think SAS provides the most functionality of any PC stat package, I don't think it is necessary or appropriate for student instruction. I am not even sure that my |STAT is appropriate for students, given that there is no student manual, only a users manual. >I'd like to see more good statistical software for PCs--I don't think that >we should sit back and let SAS Institute do it all (good as they are). We >need innovative approaches like Gary Perlman's UNIX|STAT and Analytical >Engineering's PowerStat. As PC's get faster, and mass storage cheaper, SAS will become more palatable. I hope there is enough competition to motivate SAS to get better and cheaper. Stat packages seem to be a good candidate for commodity pricing sometime in the future. -- Gary Perlman Wang Institute Tyngsboro, MA 01879 (617) 649-9731 UUCP: decvax!wanginst!perlman CSNET: perlman@wanginst
perlman@wanginst.UUCP (07/14/86)
I received several requests for information about my |STAT programs, enough that I have decided to again post my blurb on it. For current users, there is nothing new except for dsort (multiple key sort), which was added mostly because the MSDOS sort is too dumb to sort numerically. UNIX|STAT 5.2 Gary Perlman DATA MANIPULATION & ANALYSIS PROGRAMS FOR UNIX and MSDOS UNIX|STAT is a set of over 20 data manipulation and analysis programs developed at the University of California, San Diego and at the Wang Institute of Graduate Studies by Gary Perlman. The programs are designed with the UNIX philosophy that individual programs should be designed as tools that do one task well and produce output suitable for input via pipes to other programs. Interactive use is supported in the command line interpreter which also provides a programming language for complex analyses. Typical usage involves a pipeline of transformations of data followed by input to an analysis program, summarized schematically by: INPUT DATA | TRANSFORM | ANALYSIS | OUTPUT RESULTS Functions built into many statistical packages (e.g., graphics and editing) are not re-invented in UNIX|STAT which delegates such responsibility to standard tools. DATA MANIPULATION PROGRAMS: abut join data files colex column extraction dm column oriented data manipulator dsort multiple key data sorting filter maketrix create matrix type file from free-form file perm permute line order probdist probability distribution functions ranksort rank order columns repeat repeat a pattern or file reverse reverse lines, columns, and characters series generate a series of numbers transpose transpose matrix type file validata verify data file consistency DATA ANALYSIS PROGRAMS: anova multi-factor analysis of variance calc interactive algebraic modeling calculator contab contingency tables and chi-square desc descriptions histograms, frequency tables dprime signal detection d' and beta calculations oneway one-way anova and t-test with error-bar plots pair paired data statistics, regression, plots regress multivariate linear regression and correlation stats report summary statistics ts time series analysis and plots FEATURES: easy to use programs (negligible training period) simple input formats (free format field oriented) used in pipelines with other utilities (sort, print) flexible data manipulation several simple lineprinter plotting options data validation (range and type checking) documentation includes online manual entries, Handbook consistent option conventions with online help runs on any UNIX System (V6, V7, 2.8BSD, 4BSD, System V, etc.) runs on MSDOS 2.0 and 3.0 with 96K (IBM, Wang, AT&T, etc.) usually less than a few seconds per analysis liberal copyright (but can't be distributed for gain) in use at hundreds of MSDOS and UNIX sites since 1980 WARRANTY AND DISTRIBUTION CONDITIONS: Carefully read the following warranty and conditions. If you do not find them acceptable, you should not order or use UNIX|STAT. Warranty. UNIX|STAT is provided ``as is'' and without any warranty expressed or implied. There is no claim that UNIX|STAT is suited to your needs, nor that it will run on your particular hardware/software configuration. The availability of and programs in UNIX|STAT may change without notice. Copyright Conditions. You may make copies of the programs, in source, object, or binary form, and you may make copies of the ONLINE pre-formatted manual entries, provided that there is no material gain involved, and provided that the information in this flyer accompanies every copy. You may not modify the source code for any purposes other than getting the programs to work on your system. You may not distribute any modified source code to any sites other than your own. Any costs in compiling or porting UNIX|STAT to your system are your's alone, and not any other parties. Use At Your Own Risk. All risk of loss of any kind due to use of UNIX|STAT is with you, the user. You are responsible for all mishaps, even if the programs prove to be defective. You must be trained at your expense to learn to use the programs. In particular, you are hereby given fair warning that data analysis programs make many assumptions about data, These assumptions affect the validity of conclusions you may make based on the programs. References to appropriate statistical sources are made in the UNIX|STAT Handbook and in the manual entries for specific programs. ORDERING INFORMATION: Carefully read the instructions below. Orders not following them may be be returned or even discarded. All prices include delivery and should be prepaid to G. Perlman. Checks must be in US funds, drawn on a US bank. Orders that include any terms or conditions other than those in this notice may be returned or discarded. Company purchase orders should be avoided in favor of a personal check, with a request for a receipt if needed. Orders must include a delivery mailing label acceptable to the post office, and international orders must include the country name on the label. Orders without mailing labels may not be deliverable or returnable. UNIX Version of UNIX|STAT: Price: $20 Contents: Program (C language) & Manual (troff) Source Files Space: 450K bytes for source and manuals, 450K bytes for binaries Format: 600 foot half inch 9 track mag tape, 1600 bpi tar format MSDOS Version of UNIX|STAT: Price: $15 Contents: Preformatted Manuals and Executables Space: 700K for binaries and online manuals Format: 2S/2D MSDOS 5.25 inch floppy diskettes Handbook: Price: $10 Contents: Examples, Conventions, Reference Materials, CALC & DM Manuals Format: Printed Manual (~70 pages) Printed Manual Entries: Price: $5 Contents: Typeset Versions of Manual Entries, Reference Sheets Format: Printed Manual (~30 pages) CONTACT: Gary Perlman Wang Institute of Graduate Studies Tyng Road Tyngsboro, MA 01879 USA UUCP: decvax!wanginst!perlman CSNET: perlman@wanginst NOTES: UNIX is a trademark of AT&T Bell Laboratories. MSDOS is a trademark of MicroSoft. UNIX|STAT is NOT a product of any company or organization. -- Gary Perlman Wang Institute Tyngsboro, MA 01879 (617) 649-9731 UUCP: decvax!wanginst!perlman CSNET: perlman@wanginst
tankus@hsi.UUCP (Ed Tankus) (07/14/86)
Gary and Ian both ask the following: > > >I'd like to see more good statistical software for PCs--I don't think that > >we should sit back and let SAS Institute do it all (good as they are). We > >need innovative approaches like Gary Perlman's UNIX|STAT and Analytical > >Engineering's PowerStat. > As PC's get faster, and mass storage cheaper, SAS will become more > palatable. I hope there is enough competition to motivate SAS to > get better and cheaper. Stat packages seem to be a good candidate > for commodity pricing sometime in the future. > > Ian Spence {allegra,decvax,seismo,utzoo,watmath}!utcs!utcsstat!spence > -- > Gary Perlman Wang Institute Tyngsboro, MA 01879 (617) 649-9731 > UUCP: decvax!wanginst!perlman CSNET: perlman@wanginst There are quite a few good statistics packages for the PC: Statgraphics Systat Statpac Stats-2 Statpro to name a few. All run on the PC (vanilla included, I'm sure). Systat and Statgraphics are probably the best selling. Statgraphics, Systat, Statpac, Stats-2, Statpro are all trademarks of their respective companies. -- -- Ed. Net : {noao!ihnp4!yale!}!hsi!tankus Snail: Health Systems Int'l, 100 Broadway, New Haven, CT 06511 Bell : (203) 562-2101