You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Some models are based on the mathematical derivative which is better estimated by the symetric difference $dx_i \approx \frac {x_{i+1} - x_{i-1}} 2$ rather than the first différence $\Delta x_i= x_{i+1} - x_i$.
This is speciallly important if the data does not come at regular interval. In such case you should use the approximation $$\frac {dx} {dt} \approx \frac {x_{i+1} - x_{i-1}} {t_{i+1} - t_{i-1}}$$
to compensate for the fact that the $t_{i+1}- t_i$ may vary in time.
Expected behavior
Propose to compute the symetric difference.
Propose to have an Adapative time step.
unless this is what is called change quotient in which case the label is rather ambigous, and do not allow to compute the logarithmic derivative $\frac {dx} x$ or the elasticity $\frac {\Delta x} x$.
Actual behavior
Those options are not proposed.
Steps to reproduce the behavior
Not a bug but "nice to have".
Additional info (worksheets, data, screenshots, ...)
While you are there, could you please update the documention about the "change quotient" option.
The text was updated successfully, but these errors were encountered:
Thanks for this suggestion. I agree this should be changed, preferably in a way that would let the user choose the difference operator (at least to maintain backwards compatibility with existing outputs).
As I just wrote in #278, this add-on would need a caretaker. (Also, some house-cleaning would be welcome.)
Ok. Let me some time to get my hands ons (which may take a while: the learning curve is smooth but there is a long way to climb), and I'll come back with a proposal.
Timeseries version
0.6.3
Orange version
3.36.2
Use case
Some models are based on the mathematical derivative which is better estimated by the symetric difference$dx_i \approx \frac {x_{i+1} - x_{i-1}} 2$ rather than the first différence $\Delta x_i= x_{i+1} - x_i$ .
This is speciallly important if the data does not come at regular interval. In such case you should use the approximation
$$\frac {dx} {dt} \approx \frac {x_{i+1} - x_{i-1}} {t_{i+1} - t_{i-1}}$$ $t_{i+1}- t_i$ may vary in time.
to compensate for the fact that the
Expected behavior
Propose to compute the symetric difference.
Propose to have an Adapative time step.$\frac {dx} x$ or the elasticity $\frac {\Delta x} x$ .
unless this is what is called change quotient in which case the label is rather ambigous, and do not allow to compute the logarithmic derivative
Actual behavior
Those options are not proposed.
Steps to reproduce the behavior
Not a bug but "nice to have".
Additional info (worksheets, data, screenshots, ...)
While you are there, could you please update the documention about the "change quotient" option.
The text was updated successfully, but these errors were encountered: