[R] Smallest Space Analysis (SSA) in R

[Michael Friendly]

On 9/22/2018 6:49 AM, Andrew wrote:
> Hi Michael
> 
> This looks like it could be really helpful in moving my project forwards 
> thank you.
> 
> I remember many years ago using (proprietary) software from the 
> University of Liverpool which did a nice job of allowing regions to be 
> defined, and then for the space to be rotated to obtain visual 
> inspection of relative distance from different angles. I appreciate that 
> smacof will not do that, but as long as the analysis allows for the 
> graph to be plotted and analysed, that's what's important.

You need not rely on the plots provided directly by a given package.
Just roll your own using standard plotting libraries.
Here is just the first Google hit on "R MDS 3D plot"

http://omnilogia.blogspot.com/2014/05/basic-2d-3d-multi-dimensional-scaling.html

which shows a rotating 3D plot, colored by a grouping variable.  Here is 
another:

http://whatzcookinlab.blogspot.com/2013/05/spinning-3d-mds-plot.html

The vegan and ade4 packages also have a variety of plots and related 
methods.


-Michael

> 
> Thank you again, and to all of those who responded.
> 
> Best wishes
> Andrew
> 
> 
> On 21/09/18 14:07, Michael Friendly wrote:
>> Smallest space analysis (SSA) is just the name given to software 
>> developed by Guttman & Lingoes around the time the various versions
>> of multidimensional scaling were being developed.  Call it Israeli MDS
>> or Falafel MDS if you prefer. The reason you encountered it in your
>> course is presumably that the instructor was trained in that.
>>
>> There are several variants of MDS-like algorithms for embedding
>> points representing objects in a space, using data representing
>> similarities or distances among objects -- metric (cmdscale)
>> and non-metric (MASS::isoMDS), using only rank order information, and 
>> a variety of
>> measures of goodness-of-fit ("stress").  I don't recall the details
>> of the SSA programs, but that should matter little conceptually.
>>
>> The smacof package offers the widest array of possibilities.
>>
>> -Michael
>>
>>
>> On 9/19/2018 7:00 AM, Andrew wrote:
>>> Hi
>>>
>>> As part of my forensics psych course, we have been introduced to
>>> Guttman's smallest space analysis (SSA). I want to explore this approach
>>> using R, but despite finding some queries on the web about this same
>>> thing, have yet to find any answers. The MASS package doesn't seem to do
>>> the job, and the only thing I have been able to find is some proprietary
>>> software HUDAP  (Hebrew University Data Analysis Package) which may/ not
>>> be compatible with R (or GNU/Linux for that matter).
>>>
>>> Does anyone have information on how to do SSA using R?
>>>
>>> Many thanks
>>>
>>> Andrew
>>>
>>>
>>>     [[alternative HTML version deleted]]
>>>
>>
>>
>