Announcement

Collapse
No announcement yet.

Suicide bombing and support amongst muslims

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • #16
    Tanzania seems to have a marked increase as insurgency grows in Zanzibar.

    Comment


    • #17
      Originally posted by Double Edge View Post
      its not that.

      Egypt has a weird anomaly where it reduces from 28% in 2006 to 8% the year after, reduced by a factor of 3 only to creep up back to 28% by 2011. As extreme a swing such as this can't be right.

      Same with Turkey. Drops from 16% to 3% this time a year after Egypt then goes galloping up six fold to 2014.
      Ok, lets consider a few things while playing the devil's advocate.

      The first point I want to make is very important. The table has about 80 calculations over an 11 year period for 14 regions. The margin of error is high. And anomalies should be present in about 5% of the case as determined by the statistical test. So when you look through the table and pull out 2 examples you are possibly just illustrating the natural deviation from reality that one should expect to see in such a dataset. For example if you had picked out the trends at random you may have chosen the 2008-2011 Turkey period.

      That said, lets take the Turkey example first.

      2007 - 16
      2008 - 3
      2009 - 4
      2010 - 6
      2011 - 7
      2012 - no data
      2013 - 16
      2014 - 18

      Margins of error is 4.5% for 2014 for Turkey, I am going to round that up to 5%. This means if you repeated the test in 2014 during the same time period in Turkey, the statistical test predicts that you would get a result between 13-23 (5 points either side of 18) on 95 out of 100 occasions. This is the prediction by the test and ergo, the researchers. The only true way to ascertain the real number is to ask the millions of muslims in Turkey over the brief time period in question.

      That means the actual result for 2007 could be 9-21 with 95% certainty and a number approaching zero to 8% in 2008, assuming the margins of error were similar. The difference between 8 and 9% doesn't represent a big as a anomaly as simply comparing the averages without factoring in the margins of error. Especially when you consider 2 other points, first, the point I made first about the expectation of anomalies and the fact you selected them out over alternatives, and second, that there may be possibly be jumps in public opinion that are not gradual, but erratic instead.

      In addition the 2008 to 2011 turkey period are so close that it is impossible to meaingfully make a distinction between them when you consider the margins of error, but their consistency does indicate a meaningful drop on 2007, albeit, that drop may be as dramatic as the averages indicate. Likewise, the increase in 2014 may be as significant as the averages indicate, the absence of 2012 doesnt help as it may be cloaking a gradual rise (although the size of the margins of error make it impossible to say either way). Further years will help establish (like 2008-2011 did for the 2007 data) if a real change has occurred, even (once again) if that change may be much smaller than simply comparing the averages between 2011 with 2014 respectively.

      Also you refer to a change 6-fold, by being selective, and also, an increase from 3-18%, is a lot different than from 10 to 60%, so 6 fold is misleading in the terms of analysing a dataset, and either way, the nature of the data does not allow for such a conclusion to be a meaningful one.


      Originally posted by Double Edge View Post
      its not that.

      Egypt has a weird anomaly where it reduces from 28% in 2006 to 8% the year after, reduced by a factor of 3 only to creep up back to 28% by 2011. As extreme a swing such as this can't be right.
      The margins of error for egypt's data 2014 is 4.3%. Again I will make the same assumption for other years.

      2006 -28 (24-32)
      2007 -8 (4-12)
      2008- 13 (9-13)
      2009 - 15 (11-19)
      2010 -20 (16-24)
      2011- 28 (24-28)

      All the same applies as before, the test can still be considered robust despite the fact that the numbers will be outside the stated range (in brackets) on 5% of the occasion, so expect anomalies, and the changes may not be as dramatic when you consider the margins of error. Still the data does hint at a drop of some nature between 2006 and 2010 before rising again.

      Conclusions, unless the changes in opinion are dramatic, in percentage points not magnitude (like 6 fold), the dataset is not going to detect it with a high level of certainty. We have to make some assumptions on the nature of trends that we should expect to find, to what extent do such views truly vary on an annual basis, can trends be erratic or can they only be gradual? Clearly, the sample size is small and the data is limited. It probably is safe to say that the dataset is useful on a number of levels...

      1. for detecting large variation over time
      2. for establishing broadly the level of support in a country and broadly comparing countries
      3. for establishing that the level of support is not as low as some people may have reasonably considered before the research, or subsequently as they are unaware of it, that support may have been less than 1% in the muslim community. Before I saw this research I had expected the numbers to be no higher than a 1-2% in nearly all countries
      4. for countering the occasional (hopefully) bigot who thinks all muslims, an overwhelming majority, are violent, or that it is a potentially valuable statement to call Islam the religion of war, or peace for that matter, with the assumption that such a statement could ever be a meaningful one in the first place.

      The dataset is limited, but not useless. Personally, I think its very important.

      Comment

      Working...
      X