 |
 |
Anyone here able to advise on a stastical method that essentially, converts correlations into transects. The idea of transects, e.g. of a city, is attractive to geographers and some of us have probably used it as a fieldwork teaching tool - you get a bunch of students (assuming your university is located in a city with a good metro network), get them all day passes and they have to alight at e.g. every 2nd stop from end to end of a line and make notes on the types of housing, shops, businesses, people etc they see around them. Now correlations - of course on their own they prove little - but in ‘Food access diet and health in the UK: anempirical study of Birmingham’, BritishFood Journal, Vol. 114, Issue 4, (2012),these were used to create a transect. For example, for the 200+ MLSOA census areas of Birmingham and Wolverhampton, there was quite a strong +ve correlation between obesity and unemployment levels. Not suprising, obesity is strongly associated with poverty, and associated indicators such as level of County Court Judgements, and most strongly, level of qualifications attained, negatively, i.e. an MLSOA with large numbers of people with degree-level qualifications or higher was very likely to have low levels of obesity. That only tells you so much. To convert to a transect, re-do the correlation between % obese and % unemployment with the 10% of MLSOAs with the highest unemplpyment shaved off. Then repeat with the next 10% shaved off, and so on. Then repeat with the lowest 10% unemployment shaved off and so on. You now have 19 different correlation values, effectively a social transect from very low unemployment areas to very high ones. In general one can test for variable A against B as one moves from high levels of indicator C to low levels of C; B and C can be the same socio-economic indicator or different. This exercise was interesting in that, as one appraoched the very high unemployment level MLSOAs, the correlation between obesity and unemployment became much less positive and then became negative. The narrative appeared to be that, being unemployed in a low unemployment area created a propensity to be obese, BUT, being unemployed in a high unemployment area gave a LOWER propensity to be obese - other data suggested that in less affluent, high unemployment districts, the unemployed could access street markets and had the time to access fresh produce cheaply and cook it; whereas in such areas minimum wage jobs did not provide much more income than being on Benefits. In other more affluent areas being unemployed did not give easy access to such markets, and those employed did enjoy higher wages and could eat more healthily than the local unemployed there. My query is, has anyone here attempted similar methodology, did it prove useful, what other statstical techniques did they use to reinforce findings this way? Dr Hillary J. Shaw Director and Senior Research Consultant Shaw Food Solutions Newport Shropshire TF10 8NB www.fooddeserts.org