Group data based on direct
显示 更早的评论
I have a table with values and I want to group them in sectors. Is it possible to do it automatically either than manually?
4 个评论
Adam Danz
2023-1-8
Grouping a timetable by month is not difficult.
Grouping by a degree variable is also not difficult.
That's about as specific as I can get with the limited info. Also, it's not clear how you want to group by two different values (month and degrees, what are sectors?). Examples are worth 1000 words 🙂
Walter Roberson
2023-1-8
Grouping wind direction by strict sector is often misleading because the angle for wind just barely east of north is so different than the angle for wind just barely west of north. Calculating averages using such numbers is especially messed up unless you are careful.
Grouping angular values
This demo works with degrees but it's the same process for radians. It assumes angular data are between 0 and 360. If that's not the case in your data, wrap your data to 360 (or to 2pi).
Given the number of bins nBins this computes equispaced bin edges, with a bin centered at 0 deg, and then computes the bin number for each angular data point. The bin centered at 0 is delt with separately.
Grouping isn't too difficult. However, depending on what you plan to do with these group values, you could run into trouble that @Walter Roberson explained. Phase wrapping is a potential solution to this problem (I'm looking at you bin #1).
theta = [rand(1,30)*360]; % vector of angular data [0,360] deg
nBins = 12; % select number of bins
% Compute edges
w = 360/nBins;
edges = w/2:w:(360-w/2);
% Bin data
bin = discretize(theta, edges)+1;
bin(theta>=edges(end) | theta<edges(1)) = 1; % bin at 0
% Plot results, show theta values, bin edges, and bin ID
polaraxes('ThetaZeroLocation','top',...
'RTickLabel','', ...
'ThetaDir','ClockWise')
hold on
polarplot(theta*pi/180, 1, 'ro')
polarplot([1;1].*edges*pi/180, [0;1], 'k-')
text(theta*pi/180, ones(size(theta))-.1, string(bin), ...
'HorizontalAlignment','Center', 'VerticalAlignment', 'middle')
grid off
Walter Roberson
2023-1-9
See also the discussion at https://www.mathworks.com/matlabcentral/answers/128381-how-to-calculate-mean-wind-direction
采纳的回答
WindDir = rand(100,1)*360;
edgev = linspace(0, 360, 9);
[N,Edges,Octant] = histcounts(WindDir, edgev);
WindOctants = table(WindDir,Octant)
The first output are the number of counts in each octant (bonus information), and the third is the respective octant.
EDIT — (8 Jan 2023 at 16:30)
Added table.
.
11 个评论
My pleasure!
Perhaps something like this —
WindDir = rand(100,1)*360;
edgev = linspace(0, 360, 17);
edgem = [edgev(1:end-1); edgev(2:end)].';
ctrs = mean(diff(edgev))/2 + edgev(1:end-1);
[N,Edges,Decihexctant] = histcounts(WindDir, edgev);
WindDecihexctants = table(WindDir,Decihexctant)
WindDirCounts = table(edgem(:,1),edgem(:,2),N(:), 'VariableNames',{'Start','End','Count'})
figure
bar(ctrs, N)
xlabel('Direction (°)')
ylabel('Counts')
xticks(ctrs)
axis('padded')
That only required a couple tweaks to my previous code.
.
Adam Danz
2023-1-9
Thank you!
Possibly something like this —
DT = datetime('01-Jan-2022') + caldays(0:365).';
WindDir = rand(size(DT))*360;
TT = timetable(DT,WindDir)
MM = month(TT.DT);
T = timetable2table(TT);
edgev = linspace(0, 360, 17);
edgem = [edgev(1:end-1); edgev(2:end)].';
ctrs = mean(diff(edgev))/2 + edgev(1:end-1);
[N,Edges,Decihexctant] = histcounts(WindDir, edgev);
WindDirCounts = table(edgem(:,1),edgem(:,2),'VariableNames',{'Start','End'});
for k = 1:12
Month = T(MM==k,:);
[N,Edges,Decihexctant] = histcounts(T.WindDir(MM==k,:), edgev);
Nc{k} = N;
Dch{k+2} = Decihexctant;
WindDirCounts{:,k+2} = N(:);
MMM{k} = month(Month{1,1},'shortname');
end
WindDirCounts.Properties.VariableNames = {'Start','End','Jan','Feb','Mar','Apr','May','Jun','Jul','Aug','Sep','Oct','Nov','Dec'}
I thought about using splitapply or groupsummary for this, however it appears to be too complicated for those approaches.
My initial effort was to use the ‘MMM’ cell array to define the months for ‘WindDirCounts.Properties.VariableNames’ however that wouldn’t work and I finally gave up on that (accounting for the time delay) and just typed everything out. (That should be easier!)
.
Star Strider
2023-1-9
Thank you!
Star Strider
2023-1-10
There are two synchronize functions that may do what you want. I linked to the one here that looks most promising. (I never used either version of this function, so I have no experience with them.)
Star Strider
2023-2-7
I do not understand.
The current code uses direction, not velocity. However if you want to use velocity instead of direction, make the appropriate substitutions for velocity instead of quadrant. (It would likely not be possible to combine both in ome table.)
Possibly something like this —
DT = datetime('01-Jan-2022') + caldays(0:365).';
WindVel = lognrnd(log(5),log(3),size(DT));
TT = timetable(DT,WindVel)
MM = month(TT.DT);
T = timetable2table(TT);
edgev = linspace(0, 100, 101);
edgem = [edgev(1:end-1); edgev(2:end)].';
ctrs = mean(diff(edgev))/2 + edgev(1:end-1);
[N,Edges,Decihexctant] = histcounts(WindVel, edgev);
figure
bar(ctrs, N)
grid
xlabel('Velocity')
ylabel('Counts')
WindVelCounts = table(edgem(:,1),edgem(:,2));
for k = 1:12
Month = T(MM==k,:);
[N,Edges,Decihexctant] = histcounts(T.WindVel(MM==k,:), edgev);
Nc{k} = N;
Dch{k+2} = Decihexctant;
WindVelCounts{:,k+2} = N(:);
MMM{k} = month(Month{1,1},'shortname');
end
MMMv = [MMM{:}];
WindVelCounts.Properties.VariableNames = {'Start','End', MMMv{:}}
Change the second and third parameters in ‘edgeV’ to reflect the actual data, if necessary. For 1 m/s increments, the third parameter must be 1 greater than the second parameter, and both would be integers (the third must be an integer), for best results. (This uses ‘MMM’ to define the variable names in the table.)
.
As always, my pleasure!
I create the data with those lines. I do not remember any data files being posted with wind velocity (or direction), and I need data to test my code. Use the actual data instead of my synthetic data. (The only actual data I have seen thus far as been what I assume to be air temperature.)
If you post the data (with wind velocity, wind direction, temperature, and anything else necessary), I will update my code to import the files and process those data.
EDIT — (8 Feb 2023 at )
LD = load(websave('T','https://www.mathworks.com/matlabcentral/answers/uploaded_files/1289030/T.mat'))
% DT = datetime('01-Jan-2022') + caldays(0:365).';
% WindVel = lognrnd(log(5),log(3),size(DT));
TT = LD.T
% return
MM = month(TT.date_time);
T = timetable2table(TT);
Vmax = ceil(max(TT.Vel_avg)) % Maximum Velocity In File
edgev = linspace(0, Vmax, Vmax+1); % Edge Vector
edgem = [edgev(1:end-1); edgev(2:end)].'; % Bin Limits (For Output Table)
ctrs = mean(diff(edgev))/2 + edgev(1:end-1); % Bin Centers
[N,Edges,mps] = histcounts(TT.Vel_avg, edgev); % Call 'histcounts'
figure
bar(ctrs, N)
grid
xlabel('Velocity')
ylabel('Counts')
WindVelCounts = table(edgem(:,1),edgem(:,2)); % Initialise Table
for k = 1:12
Month = T(MM==k,:);
[N,Edges,mps] = histcounts(T.Vel_avg(MM==k,:), edgev); % Call 'histcounts¹
Nc{k} = N;
Dch{k+2} = mps;
WindVelCounts{:,k+2} = N(:);
MMM{k} = month(Month{1,1},'shortname');
end
MMMv = [MMM{:}];
WindVelCounts.Properties.VariableNames = {'Start','End', MMMv{:}}
figure
TL = tiledlayout(4,3);
for k = 1:12
nexttile
bar(ctrs,WindVelCounts{:,k+2})
grid
ylim([0 1.8E+4])
title(MMM{k})
end
title(TL,year(TT.date_time(1,1)))
.
Star Strider
2023-2-8
As always, my pleasure!
I thought I could do this with one of the table functions, however extracting and transposing proved to be easiest —
LD = load(websave('T','https://www.mathworks.com/matlabcentral/answers/uploaded_files/1289030/T.mat'))
% DT = datetime('01-Jan-2022') + caldays(0:365).';
% WindVel = lognrnd(log(5),log(3),size(DT));
TT = LD.T
% return
MM = month(TT.date_time);
T = timetable2table(TT);
Vmax = ceil(max(TT.Vel_avg)) % Maximum Velocity In File
edgev = linspace(0, Vmax, Vmax+1); % Edge Vector
edgem = [edgev(1:end-1); edgev(2:end)].'; % Bin Limits (For Output Table)
ctrs = mean(diff(edgev))/2 + edgev(1:end-1); % Bin Centers
[N,Edges,mps] = histcounts(TT.Vel_avg, edgev); % Call 'histcounts'
figure
bar(ctrs, N)
grid
xlabel('Velocity')
ylabel('Counts')
WindVelCounts = table(edgem(:,1),edgem(:,2)); % Initialise Table
for k = 1:12
Month = T(MM==k,:);
[N,Edges,mps] = histcounts(T.Vel_avg(MM==k,:), edgev); % Call 'histcounts¹
Nc{k} = N;
Dch{k+2} = mps;
WindVelCounts{:,k+2} = N(:);
MMM{k} = month(Month{1,1},'shortname');
end
MMMv = [MMM{:}];
WindVelCounts.Properties.VariableNames = {'Start','End', MMMv{:}}
VN = WindVelCounts.Properties.VariableNames;
WindVelCountsTransposed = array2table(table2array(WindVelCounts(:,3:end)).', 'RowNames',VN(3:end));
WindVelCountsTransposed.Properties.VariableNames = compose('[%d-%d]',edgem)
% figure
% TL = tiledlayout(4,3);
% for k = 1:12
% nexttile
% bar(ctrs,WindVelCounts{:,k+2})
% grid
% ylim([0 1.8E+4])
% title(MMM{k})
% end
% title(TL,year(TT.date_time(1,1)))
All the original information is still there in case you need it. Plotting it using bar3 might be another option, in addition to tiledlayout.
.
Star Strider
2023-2-23
As always, my pleasure!
更多回答（0 个）
类别
在 帮助中心 和 File Exchange 中查找有关 Data Type Identification 的更多信息
Community Treasure Hunt
Find the treasures in MATLAB Central and discover how the community can help you!
Start Hunting!
Translated by 
