[Yourtopia] Fwd: Normalised values

[Velichka Dimitrova]

Thank you Giacomo, I appreciate this input, feel more secure about using
re-scaling now, I am answering cc-ing the list, as I think you missed them
replying only to me.



---------- Forwarded message ----------
From: Giacomo Luchetta <giacomoluchetta at gmail.com>
Date: Thu, Mar 22, 2012 at 8:39 AM
Subject: R: [Yourtopia] Normalised values
To: Velichka Dimitrova <velichka.dimitrova at okfn.org>


Dear Velicka, dear all, ****

** **

thanks for having added me to the mailing list of this great project. For
those who do not know me, my name is Giacomo Luchetta and I am a phd
student in law and economics at LUISS University and researcher in
regulatory affairs at the Centre for European Policy Studies.****

** **

If I may give a short contribution, please find attached a “short guide” to
normalization from the JRC of the European Commission and OECD. Currently
you are using re-scaling, which is one of the most common methods (together
with the z-score) and needs not any assumption on the underlying
distribution of data.****

** **

In my experience, given the type of indicator, you may want to use
re-scaling, z-score (as the World Governance Indicators do), or possibly
ranking (as Doing Business does). There is not a correct choice, as long as
the choice is transparent, meaning that there is hardly any rule telling
that for certain data you have to use a certain normalization method. Maybe
you would like to have a pilot round of surveys and apply a sensitivity
analysis to the normalization method, to see whether its choice has any
effect on final results. It is extremely rare that composite indicators, if
not exceptionally complex, are sensible to the normalization method, but
you may want to be sure in any case!****

** **

Have a good day****

giacomo****

** **

*Da:* yourtopia-bounces at lists.okfn.org [mailto:
yourtopia-bounces at lists.okfn.org] *Per conto di *Velichka Dimitrova
*Inviato:* giovedì 22 marzo 2012 00:11
*A:* yourtopia at lists.okfn.org
*Oggetto:* [Yourtopia] Normalised values****

** **

Hi all,****

** **

I had a good chat today with David about the Yourtopia dataset (
http://thedatahub.org/dataset/yourtopia-italy) and he raised the question
about the right kind of normalisation for the values of the indicators. The
one we used today (David helped me to do pivot tables in Excel which I find
really useful), uses the Human Development Index formula (
http://en.wikipedia.org/wiki/Human_Development_Index), namely index_x= (x -
min(x)) / (max(x) - min(x)).****

** **

But would this be the right kind of normalisation and under what
assumptions? The majority of the indicators are not normally distributed.***
*

** **

I guess for Yourtopia 2 in the Italian case we're not going to have pie
charts where decreasing the value of one indicator would automatically
increase the value of other indicators. We have 21 regional data series
within 7 categories - so probably the user should do the quiz by deciding
how important each one is on a scale from 1 to 5, but how does this relate
to the normalised values in the dataset? Do we want to have the region
which has the highest average over the whole period 2000-2010?****

** **

The Yourtopia team could tell us how this was done in the first version as
I am not aware of the technical details. We should discuss what makes sense
when having regional time series - it's a different picture.****

** **

@David, do you have any ideas and questions?****

 ****

-- 
Velichka Dimitrova
Open Economics Working Group Coordinator
Open Knowledge Foundation****
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.okfn.org/pipermail/yourtopia/attachments/20120322/3eb56c8b/attachment-0001.htm>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: normalisation.pdf
Type: application/pdf
Size: 296485 bytes
Desc: not available
URL: <http://lists.okfn.org/pipermail/yourtopia/attachments/20120322/3eb56c8b/attachment-0001.pdf>