Bookmark and Share

Notice: On March 31, it was announced that Statalist is moving from an email list to a forum. The old list will shut down at the end of May, and its replacement, statalist.org is already up and running.


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: st: which statistical analysis to use


From   David Hoaglin <dchoaglin@gmail.com>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: which statistical analysis to use
Date   Thu, 19 Apr 2012 08:04:55 -0400

Nick,

If a substantial number of the 27 skills were scored as 0 by all
companies, one could (as you suggested earlier) set those skills
aside.  That might simplify the analysis.  (The analysis would be
conditional, because a skill with an observed score of 0 could have a
positive probability of receiving a positive score, but the impact
seems small when all 300+ companies gave it a 0.)  An look at which
sets of skills received nonzero rankings should be the first step.

The constraint is that an individual company can (assuming it can't
give tied scores) assign only one 1, only one 2, etc.  To analyze the
data as measurements, we would need a question that allowed a company
to assign any score it wished to a particular skill.  One might, for
example, ask, "On a scale of 0 to 100, how important is this skill to
your company?"  Research may have shown that such a question tends not
to produce good, reproducible data; and it is probably more effort for
the company than the actual ranking, leading to higher rates of
nonresponse, etc.  The data collection has to face many practical
problems.

David

On Thu, Apr 19, 2012 at 7:27 AM, Nick Cox <njcoxstata@gmail.com> wrote:
> I think I agree to all. The ranking literature I have seen by the
> authors you quoted earlier contains some really cute methods for the
> case in which all the ranks are distinct (no ties). Then the ranks are
> permutations of the integers 1 up and group theory and goodness knows
> what lead to some very smart analyses. Here we are at the opposite
> end, in which tieing is massive. I wouldn't expect much from that
> direction.
>
> As you say, there are constraints here, so there can only be so many
> 1s, so many 2s and so forth. That constrains the companies, but it
> doesn't constrain the skills except indirectly, as I understand it.
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/


© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   Site index