I have x-cordinate graphs plotted against time of a particular data. For more accuracy the data is taken 3 times n hence I have a set of 3 graphs. The graphs obtained are similar but not up to the accuracy of even 90%..and there is no method of each time taking the data with more accuracy. Now I need to set a single standard graph that incorporates the features of all the 3 (sample graphs obtained) with more accuracy. Please anybody suggest me the method to do so.
Asked
Active
Viewed 285 times
-1
-
2Do you have the values of the originally acquired data? I don't really understand what you're trying to do. Are you looking for some kind of bias removal? Would calculating the mean of all 3 samples be a solution? – Deve Jun 28 '12 at 12:16
-
Plot the mean for each time point? – Dan Jun 28 '12 at 13:44
-
yes i have the values for all the three sample data but taking mean might give wrong results for the case in which 2 sample sets are similar and one quite different..any other solution than taking mean..? – Nehal Jun 28 '12 at 14:13
-
1How can you be sure that the one sample set you consider beeing "quite different" isn't the sample set that represents the actual data the best? The two "similar" sets could both be wrong. 3 samples isn't a good foundation for a statistically accurate estimation of the actual signal. – Deve Jun 28 '12 at 15:31
-
1Why not use a boxplot? Why not just plot three overlapping lines? – slayton Jun 28 '12 at 15:53
-
whats the solution if I take 5 sample datas instead of 3? and is there a better option than taking mean? – Nehal Jun 29 '12 at 07:15
-
Mean is very sensitive to outliers. Median is more tolerant to these. – bdecaf Jun 29 '12 at 12:28
1 Answers
1
An example of using mean/median:
%# data and noisy estimates
t = 1:100;
x = cumsum(rand(size(t))-0.5);
x1 = x + randn(size(x))*0.7;
x2 = x + randn(size(x))*0.9;
x3 = x + randn(size(x))*1.2;
%# plot estimates
clf
subplot(211)
plot(t,x1, t,x2, t,x3)
legend({'estimate 1','estimate 2','estimate 3'})
%# mean/median across estimates
subplot(212)
plot(t,mean([x1;x2;x3]), t,median([x1;x2;x3]))
hold on, plot(t,x, 'm:', 'LineWidth',2), hold off
legend({'mean','median','actual'})
If you are concerned about outliers, see this for ideas.
