Daniel Siefman have you got the answer?I'm looking for this too

A bit late, but seeing how there is no answer, I just answered a similar question, and it might be useful for others, thought I'd give a shot.

The number of trajectories is typically between 4 to 10, according to Franczyk [1] and Satelli et al. [2]

Ruano et al. [3] details an approach to select an optimal number of trajectories. The approach is implemented in SALib [4], a Python package for sensitivity analysis (full disclaimer: I help maintain the package)

[1] Article Using the Morris sensitivity analysis method to assess the i...

[2] Article Sensitivity Analysis in Practice: A Guide to Assessing Scien...

[3] Article An improved sampling strategy based on trajectory design for...

[4] https://github.com/SALib/SALib

Hi Daniel,

Yes, I am also late to response your question (simply didn't see it before). I hope you would still find something useful for this topic in my recent paper on use of the Morris method in spectroscopic ellipsometry:

Article Parametric sensitivity analysis as an essential ingredient o...

Best wishes!

Information in Campolongo (2007) [1]:

'The EE method is based on the construction of r trajectories in the input space, typically between 10 and 50.'

There are also options in R (package: sensitivity) implemented to choose the 'best" trajectories out of a pool on possible trajectories (like the option from Ruano in Python, mentioned from Takuya Iwanaga) if you're using R and not Python.

Actually, I also was wondering how the grid.jump and the number of trajectories interact. You can find at least a recommendation/hint in Campolongo (2007) regarding the step size:

'A convenient choice for the parameters p and Δ is p even and Δ equal to p/[2(p − 1) ... where Δ is a value in {1/(p − 1),…,1 − 1/(p − 1)}, p is the number of levels.'

[1] Article An effective screening design for sensitivity