Plot a heatmap using small size data

Question

Raihan 2024-7-25

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/2140236-plot-a-heatmap-using-small-size-data

回答： Jatin Singh 2024-8-6，8:59

I have 20 x 3 data points: 10 different subjects with different heights and BMIs. Each subject has its own injury probability value.

I would like to know if it is possible to plot these into a heatmap. I tried to plot it but encountered NaN values at the boundaries.

I expect to plot something like this, which is from a previous study where the dataset is much larger.

If a heatmap is not possible, could you suggest an alternative way to visualize the relationship between these variables? Thank you.

1 个评论
显示 -1更早的评论隐藏 -1更早的评论

Umar 2024-7-25

Hi Raihan,

Since you mentioned that you are having trouble with heatmap plot,consider creating a scatter plot matrix in Matlab to visualize the relationship between heights, BMIs, and injury probabilities without NaN issues. I hope this will help resolve your issue.

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Jatin Singh 2024-8-6，8:59

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/2140236-plot-a-heatmap-using-small-size-data#answer_1495231

在 MATLAB Online 中打开

Hi,

The problem with encountering NaN values generally comes when the data size is small. In this scenario, to get a better visualization we might want a bigger data size.

Data interpolation is a technique that can be used to estimate values within the range of known data points. In MATLAB this can be done using the “griddata” method which is used to interpolate 2-D and 3-D data.

The “griddata” function returns a grid data which contains interpolated data using the known data points but can still contain NaN values which are out of the scope of the known data points, to fill these values we can use “fillmissing” function.

Kindly refer to the below documentations of “griddata” and “fillmissing” to get a better understanding:

https://www.mathworks.com/help/releases/R2023b/matlab/ref/griddata.html

https://www.mathworks.com/help/releases/R2023b/matlab/ref/fillmissing.html

As “heatmap” function is designed for 2D matrix data, “imagesc” function can be used to plot and visualize grid data in this case. Kindly look at the below example code to plot a heatmap with small data size and use of “griddata” and “fillmissing” for data interpolation.

% define dataset 
height = [170, 180, 175, 160, 165, 155, 172, 168, 177, 182]; 
BMI = [22, 25, 24, 20, 23, 21, 26, 22, 27, 25];  
prob = [10, 30, 20, 5, 15, 10, 35, 20, 40, 30];  
% Define the grid for height and BMI 
height_grid = linspace(min(height), max(height), 50); 
BMI_grid = linspace(min(BMI), max(BMI), 50); 
[H, B] = meshgrid(height_grid, BMI_grid); 
% Interpolate the injury probability values onto the grid 
prob_grid = griddata(height, BMI, prob, H, B, 'linear'); 
% Fill NaN values using linear interpolation 
prob_grid = fillmissing(prob_grid, 'linear'); 
% Clip the interpolated values to be within the range [0, 1] 
prob_grid = max(0.05, min(100, prob_grid)); 
% Create the heatmap using imagesc 
figure; 
imagesc([min(height), max(height)], [min(BMI), max(BMI)], prob_grid); 
% Correct the Y-axis direction 
set(gca, 'YDir', 'normal'); 
title('Probability Heatmap'); 
xlabel('Height(cm)'); 
ylabel('BMI(Kg/m^2)'); 
% Apply the custom colormap 
colormap(jet);  
colorbar; 