On Mon, Sep 13, 2010 at 8:02 AM, Marc Schwartz <marc_schwa...@me.com> wrote:
>
> On Sep 13, 2010, at 8:59 AM, Joshua Wiley wrote:
>
>> If it's not possible to use their particular algorithms, does anyone
>> think it would be helpful/practical to try to write a general scoring
>> system? I imagine a function with arguments for column names, a list
>> where each element is a vector that indicates the numbers that
>> correspond to various subscales, an argument that could handle any
>> reverse scoring, etc.
>>
>> I am willing to have a go at this if people think it would be
>> worthwhile (read: if someone wiser than me thinks it is not a waste of
>> time).
>>
>> Josh
>
> It's not clear to me what you are proposing.
>
> The SF-* instruments are validated scoring systems, that have been
> demonstrated to correlate to quality of life and in turn, to healthcare
> resource utilization and cost.
>
> Are you proposing to develop an algorithm that performs the same set of
> functions? If so, note that you would have to go through the same scoring
> system validation that originally RAND and now QualityMetric have gone
> through.
Apologies on not being clearer, I certainly did not mean develop and
validate an alternate scoring system for the SF-*. I was thinking
many scales have steps in common (like overall vs. subscale scores,
reverse coding, z-scoring items, weighting items by difficulty), and
that this could be automated in a function. Ideally, once these
initial steps were done, the user would have little left to do to
score it. I can think of a few scales in psychology where the
subscales are essentially just summed (after appropriate reversing and
scaling). This pseudo-code is an example of what I was thinking
function(dataset = mydata, variables = colnames(mydata)[1:20],
reverse.code = c(3, 6, 8, 12, 14),
subscales = list("A" = c(1:6, 11:14), "B" = c(7:10, 15:20)),
scale = TRUE/FALSE, weights = , algorithm = )
But, the more I think about it, the first part is trivial to
implement, and in the case where scoring is more complex, users are
not just going to be able to pass one line of code to the mystical
algorithm argument and be done, which answers my question whether it
would be useful.
Sincerely,
Josh
> Of course, you could use RAND's original implementation of the SF-36
> (RAND-36):
>
> http://www.rand.org/health/surveys_tools/mos/mos_core_36item.html
>
> which is in the public domain. However, there are material differences in the
> scoring systems now used by QM and the original RAND scoring mechanism, as I
> understand it, is almost never used these days.
>
> Regards,
>
> Marc Schwartz
>
>
______________________________________________
R-help@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
