Jump to content

Talk:Elementary effects method

Page contents not supported in other languages.
From Wikipedia, the free encyclopedia

I agree with the objections to this article - what reference supports the statement that this is the most commonly used SA screening method (also stated on the main Sensitivity Analysis page)? One of the primary sources itself, Campolongo et al. 2007, states that the Morris method is very useful, but still under-used. There are published papers that use this method, other than the Campolongo-Saltelli group, but I have only found a handful. The paper does support - quantitatively - its assertion that the modified u* output provides an improved measure over the originally proposed measure of u. There are many other screening methods (some similar to the elementary effects approach) used, which would be helpful to mention or link to on this page.

Rinee121 (talk) 07:04, 26 July 2010 (UTC) rinee121[reply]

I propose to merge this page with the Morris method page, since they are about exactly the same thing. Please if anyone has any comments on this, let me know. If not I will make the merge soon (probably to incorporate the Morris method page into this one, since this one is a bit more comprehensive).

Regarding the comment above, if Rinee121 knows "many other screening methods (some similar to the elementary effects approach) used, which would be helpful to mention or link to on this page", why not add them yourself? Or on the main sensitivity analysis page? WillBecker (talk) 10:40, 30 October 2012 (UTC)[reply]

Wrong number of points in each trajectory?

[edit]

The wiki wrote:
Each trajectory is composed of (k+1) points since input factors move one by one of a step Δ in {0, 1/(p-1), 2/(p-1),..., 1} while all the others remain fixed.

Based on the definition of the trajectory, the values of the index i begin from 1 and end at k. I think the trajectory is composed of k points, not (k+1) points. Smallperh45 (talk) 23:44, 1 March 2015 (UTC)[reply]

I think k+1 is correct. Imagine an experiment with k=2. You start with a point somewhere in the space, then you move a certain distance in the direction of x1, and take a new point. That's 2 points so far. But so far you have only moved in the direction of x1. You now move some distance in the direction of x2 to take the final point. So that's 3 points in total for k=2. WillBecker (talk) 08:49, 2 March 2015 (UTC)[reply]

Clarification on delta

[edit]

I have a simple question about this remark:

"In case input factors are not uniformly distributed, the best practice is to sample in the space of the quantiles and to obtain the inputs values using inverse cumulative distribution functions. Note that in this case ∆ equals the step taken by the inputs in the space of the quantiles."

Does that mean simply that in this case delta is different for each parameter? I.e. if I have two parameters with different ranges, k1 with [0,1] and k2 with [0,10], will the delta1 be 0.5 and delta2 equal 1?

Added missing references

[edit]

After including four references to sensitivity analysis review papers I removed the missing reference stub; also linked to the page of the developer of the method, Prof Max D Morris at Iowa State University.Andrea Saltelli Saltean (talk) 11:22, 20 January 2024 (UTC)[reply]