orange3
orange3 copied to clipboard
Add error bars to the scatter plot
What's your use case?
Data Science. Error Bars are crucial.
What's your proposed solution? Add possibility to add Error Bars
Are there any alternative solutions? No
Error bars of what?
We are talking about scatter plot, that is, projection of multidimensional data in two dimensions, right?
When I want to plot data X (for example Height) versus data Y (for example Weight), I would like to add errors X_error and Y_error. A X-Y scatter plot is a two dimensional plot of two dimensional data.
An example is given here: https://www.geeksforgeeks.org/use-error-bars-in-a-matplotlib-scatter-plot/
Scatter plot shows (not necessarily linear) relations between two (in principle independent) variables.
I think we must remove regression line from the Scatter Plot widget to avoid confusion. (We can keep the line that minimizes orthogonal distance from the line.) Then, we should finally add a widget for linear regression, which would show regression line, confidence intervals, and so forth, including points and error bars. There is too much linear regression-related stuff, we shouldn't burden the scatter plot with it.
Sure, but what does an error bar have to do with a regression line?
I'm sorry, I overlooked the link you posted and was somehow fixated around linear regression.
This would mean adding two more features to the existing four (shape, color, size and label): the length of a vertical and horizontal line. I'd prefer being conservative about bells and whistles (while acknowledging that one's bell might be another's wheel). Orange will never support everything that you can do in, say, matplotlib.
Not everything, but Error Bars are so crucial for so many kind of data.