Bin data and fit curve

9 次查看（过去 30 天）

显示更早的评论

MiauMiau 2017-6-20

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/345438-bin-data-and-fit-curve

评论： Farooq Muhammad 2022-3-17

采纳的回答： Walter Roberson

Hi, I have the following data:

- A vector of physical stimulus properties, something like: [0.8, 0.7, 0.56,0.4,....,0.01]

- For each of the vector's elements, the response of the subject (either "0" or "1").

What I want: I want to bin the data with the physical stimuli, and determine the percentage of "1"s responded for this bin (it would be the percentage of correct answers for the bin).

Additionally, I want a curve fittet through these percentages. I know about the hist function, but I am not sure of how I can apply it here. Thanks

1 个评论
显示 -1更早的评论隐藏 -1更早的评论

Farooq Muhammad 2022-3-17

hi all

i want the answer to simulation question.

i have the error plot of height above the sea level w.r.t to some reference system.

i want to bin the x axis and curve fit the bins using curve fitting tool or something else to know how error is behaving w.r.t height above the sea level.

help requierd.

请先登录，再进行评论。

请先登录，再回答此问题。

采纳的回答

Walter Roberson 2017-6-20

1
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/345438-bin-data-and-fit-curve#answer_271241

binedges = 0 : 0.1 : 1;            %set as appropriate
[~, ~, binnumber] = histcounts( StimulusVector, binedges );
maxbin = max(binnumber);
count0 = accumarray(binnumber, Response == '0', [1 maxbin]);
count1 = accumarray(binnumber, Response == '1', [1 maxbin]);
total_for_bins = count0 + count1;
percent_per_bin = count1 ./ max(1,total_for_bins) * 100;

If you only need the percentages and not the counts, you can make this a bit shorter:

binedges = 0 : 0.1 : 1;            %set as appropriate
[~, ~, binnumber] = histcounts( StimulusVector, binedges );
percent_per_bin = accumarray(binnumber, Response == '1', [], @mean) * 100;

To fit a curve:

x = (binedges(1:end-1) + binedges(2:end))/2;
y = percent_per_bin;

and now you can fit your curve on x and y using whatever distribution seems suitable. You might want to look inside histfit() to see how it does the fitting and plotting.

You wrote that the response is (either "0" or "1") so my interpretation is that the responses are the characters '0' or '1' . If they are instead the numeric values 0 or 1 then change the '1' in the code to plain 1

MiauMiau 2017-6-22