function [confidence, changepoint]=change_point_analysis_trend_method_with_bootstrap(input_data,bootstrap_runs,do_plotting)
%this is coded from a method by taylor, http://www.variation.com/cpa/tech/changepoint.html
%possible future additions include the consideration of a second candidate change point, to be identified at the same time as the first one

%change point analysis for variance also?

c=cumsum(input_data-mean(input_data));
[maxval_original maxind_original]=max(c);[minval_original minind_original]=min(c);
diff_original=maxval_original-minval_original;

diffs_shuffled=[];mins_shuffled=[];
for i=1:bootstrap_runs  %note that this is not a true bootstrap but rather a permutation test
    shuffled_data=shuffle_randomization(input_data);
    shuff_c=cumsum(shuffled_data-mean(shuffled_data));
	diffs_shuffled=[diffs_shuffled max(shuff_c)-min(shuff_c)];
    mins_shuffled=[mins_shuffled min(shuff_c)];
end
%close,hist(diffs_shuffled,50),hold on, plot([diff_original diff_original],[minval_original maxval_original],'r') 
%confidence=100*(length(find(diffs_shuffled<diff_original))/bootstrap_runs);    %this way is based on the size of the difference between max and min
confidence=100*(length(find(mins_shuffled>minval_original))/bootstrap_runs);    %this way is based on the depth of min
changepoint=minind_original;

if do_plotting
	plot(input_data, 'ko')
	hold on
	plot([1 changepoint],[mean(input_data(1:changepoint)) mean(input_data(1:changepoint))],'r')
	plot([changepoint+1 length(input_data)],[mean(input_data((changepoint+1):end)) mean(input_data((changepoint+1):end))],'r')
end