[nc-whois] suggestions

[Thomas Roessler <roessler@does-not-exist.org>]

May I suggest that one of you who has access to staff or executive 
assistants (or whatever) could give them the task to take the 
results we received last week from Marie Juliano, and prepare one 
Excel sheet per multiple choice question, with the sheet having as 
column labels the possible answers to the question, as row label the 
category of the respondent, and as entry in the cells the number of 
responses?

That way, we should be able to easily produce some nice pie charts 
for Ghana, with one chart by category of respondent, and one chart 
totalling them (i.e., the column sums).

Also, it would be rather easy to discern, from that set of data, 
where we have some strong common trends, and where categories of 
respondents disagree.

The questions which could reasonably be covered and evaluated that 
way are 2, 3, 4, 6, 8, 10 (the yes-no parts), 11, 12, 13, 14, 16, 
17, 18, 19.

I do believe that we could produce a preliminary report on these 
questions until Accra, when we really start off with the pie charts, 
ask (and discuss!) what conclusions can be drawn from these, and 
then go on hunting for gems among the "others" (which is faster than 
actually classifying them, because you can hunt for gems by more 
sloppy reading of answers).


We could then base the decision whether or not we investigate the 
"other" answers in detail on the the set of 300 we have all looked 
through: If some of us believe that there is information hidden in 
these answers which can't be derived from the statistical answers to 
the questions mentioned above, or if there are new answers showing 
up which are as frequent as choices given on the questionnaire, then 
we may have to look at them in a more complete manner - either by 
attacking the statistical 300 more systematically, or by attacking 
the entire set of answers received. However, we should really make a 
cost-benefit decision in that case.  ;-)



Questions 15 and the cost part of question 10 are special, since
they should probably have offered the same kind of choices:

 - registrants
 - users of search service
 - other (actually, this answer doesn't really make sense except if 
   you believe in tax payers or charities bearing the cost of such 
   services).

(Registrars (or rather: registrars' shareholders), for instance, are 
not going to pay for anything.  They want to get money back from 
their investment, so that particular answer actually means 
"registrants".)

Fortunately, question 15 allows for an easy mapping (with 4% 
answering "other", which can safely be moved to the "other" category 
above without further inspection).  Unfortunately, the cost part of 
question 10 is free-form, and indeed one of the questions where we 
need to assign each answer to one of the above classes.


Question 8.2 is also a multiple-choice question gone wrong, where 
the baskets we can use are obviously any combination of points A-I 
in q8.

Depending on how people work it may be reasonable to defer this 
stuff until after Ghana.  However, this should certainly be in the 
final report.



8.1 is a question where we need to find out about baskets by reading 
answers.


Question 12 (the "why or why not" part) is probably a place where 
it's really sufficient to look for gems - I don't believe that we'd 
really win anything for our purposes by doing a statistical analysis 
of the reasons given.


The "best way" part of questions 13 and 14 is similar.  You don't 
solve engineering problems by doing public polls.



Question 17 is another case where an approach with putting responses 
into baskets and trying to do statistical analysis on this is 
probably the only way to get some added value.


Same, of course, with question 20.

-- 
Thomas Roessler                        http://log.does-not-exist.org/