Sometimes in my calculations I got a vector with non-strict monotonic data.
E.g.: 0 0 0 50 70 90 100 or 0 10 40 60 75 85 80
As this problematic case only appears in special cases, the reason is always either some 0 values at the beginning or a decreasing at the end.
Here is my code:
interp1(sort(nanmean(TG1S(:,:,1),2)),p,0.5*max_bol)
max_bol = 95%
p = -10 -5 0 5 10 15 20
Unfortunately monotonic but not strict monotonic data also doesn't work. So sort() doesn't help (I checked all my data; sort doesn't influence my interpolation results in any case).
Is there another possibility to interpolate to get the value for 0.5*max_bol?
PS: of course also a vector like 10 20 50 50 60 70 90 would get interp1() in trouble, but this won't happen in my case, I'm sure.
For interpolation theory there are independent variable and dependent variable. The independent variables must be in increasing case.
The function interp1q is quicker than interp1 for non-uniformly spaced data and asks for monotonically disposed x vector - in (x, f(x)). The interp1 function doesn't need this.
You must just check if your point of interest "0.5*max_bol" is between min(x) and max(x).
The ordering of the elements in the x vector doesn't matter for interp1 you shoul only care about the (xi, f(xi)) pairs. And min(x) < 0.5*max_bol < max(x).
thanks for your help. interp1q excludes some problematic cases.
I just wrote a kind of fix for interp1, which picks out the needed values (maybe someone else can also use this code). This is the easiest way I found (interp1_upper_tg1s and the equivalent for TG2S are the first above 0.5*max_bol indices):
% 3D matrix size
tg1s = size(TG1S);
tg2s = size(TG2S);
tg0s = size(TG0S);
% mean values of test persons
nanmean2_TG1S = squeeze(nanmean(TG1S,2));
nanmean2_TG2S = squeeze(nanmean(TG2S,2));
% number of test persons
test_persons_G1 = tg1s(2);
test_persons_G2 = tg2s(2);
test_persons_G0 = tg0s(2);
% init interp1 fixing values (position of first vector values >=
% 0.5*max_bol_G1)
interp1_upper_tg1s = zeros(1,tg1s(3));
interp1_upper_tg2s = zeros(1,tg2s(3));
% find first value >= 0.5*max_bol_G1
for m = 1:tg1s(3)
for s = 1:(tg1s(1)-1)
if nanmean2_TG1S(s,m) > 0.5*max_bol_G1
interp1_upper_tg1s(m) = s;
break;
end
end
end
% find first value >= 0.5*max_bol_G2
for m = 1:tg2s(3)
for s = 1:(tg2s(1)-1)
if nanmean2_TG2S(s,m) > 0.5*max_bol_G2
interp1_upper_tg2s(m) = s;
break;
end
end
end
regards,
Daniel
a very simple way would have been to use the unique() function (instead of sort, for example). It deletes all elements that appear more than once. When using unique(A,'rows'), it only removes duplicate rows rather than duplicate elements. See link below
http://de.mathworks.com/help/matlab/ref/unique.html
- Badges
- Science topic