Re: st: spearman rank correlation

[Nick Cox <njcoxstata@gmail.com>]

I don't see that. Spearman correlation just establishes the strength
of monotonic relationship; it is not associated with a specific
prediction method. If you want to use -regress- for prediction the
figures of merit for assessing how well that works come from the
regression itself.

Nick
njcoxstata@gmail.com


On 19 November 2013 13:23, jean-luc morin-chesnel
<jeanluc.morinchesnel@gmail.com> wrote:
> thanks mr cox for your kind answer.
>
> well I would like to know whether past successes lead on average to futures ones
> so maybe one way would be to -reg future_score past_score, vce(cluster id)
> but is seems that using a spearman rank correlation is a more robust
> way to go (I guess)
>
> thanks again
>
> On 19 November 2013 12:50, Nick Cox <njcoxstata@gmail.com> wrote:
>> I don't see how Spearman correlation is natural for this. Perhaps you
>> should spell out how you expect to use past scores to predict future
>> scores.
>>
>> Nick
>> njcoxstata@gmail.com
>>
>>
>> On 19 November 2013 10:41, jean-luc morin-chesnel
>> <jeanluc.morinchesnel@gmail.com> wrote:
>>
>>> I have a panel dataset for N individuals over T periods (unbalanced)
>>> and I would like to test whether a good indidivual score Y over period
>>> T=0 to T=n predicts a good score over the periods T=n+1 to T=T. In
>>> other words, is performance persistent over time accross indidivuals?
>>>
>>> I have read that I could use a Spearman rank correlation test to do
>>> so, but I have to idea how to implement this with panel data in Stata
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/