dear john
If you choose route 1, MATLAB & Psychtoolbox, we'll be happy to help,
e.g. providing web pages at our site or linking to your web pages, so
that users can easily learn about and download everything they need.
best
denis
Denis Pelli
http://psychtoolbox.org/
http://psych.nyu.edu/pelli/
John Vokey wrote:
If you choose route 1, MATLAB & Psychtoolbox, we'll be happy to help,
e.g. providing web pages at our site or linking to your web pages, so
that users can easily learn about and download everything they need.
best
denis
Denis Pelli
http://psychtoolbox.org/
http://psych.nyu.edu/pelli/
John Vokey wrote:
>X-Sender: huff@...
>Date: Mon, 13 May 2002 20:52:25 -0500
>To: macpsych@...
>From: "John Vokey" <vokey@...>
>Subject: Stats packages (again) and a proposal
>
>As others have commented, stats packages for the Mac OS are
>disappearing quickly, especially those that experimental
>psychologists find (found) most useful (i.e., CLR ANOVA, SuperAnova,
>Statview, Systat) because we frequently need the following features:
>
>(1) ANOVA for mixed between and within factors, with many of both.
>(2) Simple effects analyses of interactions of between and between,
>within and within, and, most important, between and within.
>(3) Orthogonal contrasts for effects of any complexity (i.e., where
>the error-term for the effect (say an 3-way interaction) serves as
>the error-term for the desired contrasts).
>(4) less frequently:
> (a) ANCOVA
> (b) polynomial regression
> (c) multiple regression
> (d) multi-way frequency analysis
> (e) randomization and resampling tests
>
> As with many of you, I have looked for alternatives, with MacAnova
>and R figuring prominently as both are open source packages.
>Unfortunately, only R is currently available for OS X, and, although
>MacAnova is somewhat less cumbersome, neither does any of 1--3
>without much tearing of hair and gnashing of teeth, if at all.
>Certainly, not easily.
>
> So, I have a proposal. We are individually and collectively
>smarter than your average bear, virtually everyone of us has
>moderate to extensive experience programming for the Mac, if not
>more generally, many of us are quite fluent in stats-speak and the
>programming of stats functions and routines, and we know what we
>want, which most decidedly is *not* Stata or (horrors!) SPSS. So
>why not produce what we need *ourselves*? That is, why not form an
>open source collective to produce (and keep producing) a stats
>package for experimental psychologists in general, and mac
>experimental psychologists in particular. In this regard, I see
>three (not mutually exclusive) possibilities:
>
>1. As Matlab will be available for Mac OS X, produce a stats toolbox
>of Matlab macros for experimental psychologists (rather than the
>commercial one that appears to be for engineers who use, apparently,
>nothing more complicated than a simple 2-way, between-s ANOVA) to
>accompany the VisualPsychophysics toolbox that many of us use. I
>suspect many of us already have a large collection of such routines.
>As Matlab is also available for Windoze and 'nix, the macros would
>be cross-platform, but it does require that the user have Matlab,
>which is not cheap.
>
>2. Develop the package in something easily extensible (even for the
>beginner), such as Metacard. This approach, aside from making the
>interface design simple, has the added advantage that the resulting
>stack or stacks are completely cross-platform (i.e., OS X, Mac
>pre-OS X, including 68k machines, Windoze and 'nix). Furthermore,
>for those who just want to use the stacks, the *free* metacard
>starter kit will run the stacks just fine. And metacard is a great
>environment for creating and running experiments (it's hypercard on
>steroids).
>
>3. Even more generally, write the package(s) in javascript. This
>approach has the same advantages of easy extensibility and
>cross-platform capabilities as 2 (i.e., virtually any modern
>web-browser can run it), does not require special software, is
>easily modified, and is easy to design the interface for (i.e., it's
>just html). I have searched the web for such, and although there
>are many javascript stats packages available, most do little more
>than simple stats, and none come even close to the three main
>features we desire. A few, more sophisticated packages are available
>written in java, but they also don't meet the main objectives and,
>of course, are written in java---not a language for the faint of
>heart (or, imho, for anyone who wants to get anything done). Worse
>yet, they require that one has available (and is willing to use the
>inevitably buggy java interpreter).
>
> Failing that, perhaps we could design R and MacAnova macros and
>functions, and appropriate tutorials to accomplish the main features
>we need in these packages, and share them with everybody.
>
> I'm willing to coordinate such an effort, provide web-space, and
>even contribute much of the code (at least initially), but I would
>like to know whether the rest of the MacPsych community views this
>general approach as a worthwhile endeavour, and which one or more of
>the specific approaches are favoured.
>
> Please respond *to the list* <macpsych@...>, and not just
>to me (i.e., don't simply hit "reply").
>
>--
>John R. Vokey, Ph.D. |\ _,,,---,,_
>Professor /,`.-'`' -. ;-;;,_
>Department of Psychology and Neuroscience |,4- ) )-,_. ,\ ( `'-'
>University of Lethbridge '---''(_/--' `-'\_)