how are the colors assigned to the nodes in a sensitivity analysis?

[PCherubini]

I noticed that when performing a sensitivity analysis on a complex network, the colors of the nodes do not strictly follow neither the derivatives of their parameters, nor the width of the variation that the node can cause on the target node. For example, in the attached image see an almost-dark-red "confessioni successive" node, seemingly suggesting that it is quite influent in the analysys, but - in the tornado diagram - the first of its parameters that is ranked is in rank 44... (see extract of the tornado diagram below)... far below some nodes that are pink. It is not critical, but out of curiosity and if you have time to answer, how are the colors assigned to the nodes?
Thx

Sensitivity for orrb_colpevoli=Sì
Current value: 0.999664 Reachable range: [0.998368 .. 0.999847])

2: Esame_Macchia=Diluita | OR_porta_sangue_in_auto=No, Contaminazione=True, Sangue_in_auto_rilevato=UnaMacchia
Target value range: [0.999484 .. 0.999847], width=0.000362229
Parameter range: [0.81 .. 0.99], width=0.18
Current parameter value: 0.9 at definition index 40
Derivative: 0.00201219
Coeffs: a=-1.02679e-16, b=1.0647e-15, c=-1.04671e-16, d=1.06682e-15
...
4: OR_porta_sangue_in_auto=No | OR_si_macchia_di_sangue_V=Molto, OR_rimuove_macchie=Accuratamente
Target value range: [0.999702 .. 0.999614], width=8.75264e-05
Parameter range: [0.72 .. 0.88], width=0.16
Current parameter value: 0.8 at definition index 2
Derivative: -0.00053808
Coeffs: a=-1.55591e-15, b=2.21701e-15, c=-1.55591e-15, d=2.21734e-15

5: OR_porta_sangue_in_auto=No | OR_si_macchia_di_sangue_V=Poco, OR_rimuove_macchie=No
Target value range: [0.99962 .. 0.999707], width=8.75256e-05
Parameter range: [0.693 .. 0.847], width=0.154
Current parameter value: 0.77 at definition index 17
Derivative: 0.000568348
Coeffs: a=-3.5711e-18, b=9.75036e-16, c=-4.12527e-18, d=9.7579e-16

6: Tracce_altri_vestiti_VC=True | VC_uccisa_a_piano_2_=False, orrb_colpevoli=No, Tracce_compatibili_VVFF=True
Target value range: [0.999623 .. 0.999704], width=8.14339e-05
Parameter range: [0.648 .. 0.792], width=0.144
Current parameter value: 0.72 at definition index 12
Derivative: 0.000565513
Coeffs: a=0, b=9.72287e-16, c=-5.50211e-19, d=9.7301e-16

7: scontrino_non_richiesto=True | orrb_colpevoli=Sì
Target value range: [0.999626 .. 0.999694], width=6.79281e-05
Parameter range: [0.459 .. 0.561], width=0.102
Current parameter value: 0.51 at definition index 0
Derivative: 0.000659307
Coeffs: a=1.90644e-15, b=0, c=1.90644e-15, d=3.27148e-19

...

44: Confessioni_successive=True | Confessano_ORRB=True, orrb_colpevoli=No, Altre_ragioni_2=True
Target value range: [0.999692 .. 0.999636], width=5.60411e-05
Parameter range: [0.045 .. 0.055], width=0.01
Current parameter value: 0.05 at definition index 4
Derivative: -0.00560411
Coeffs: a=0, b=9.72287e-16, c=5.45247e-18, d=9.72341e-16

[shooltz[BayesFusion]]

The values displayed in the tornado are actual target posterior changes for the specified parameter spread (see the slider control in the bottom part of the tornado dialog). These are not the derivatives.

The coloring of the nodes when the sensitivity analysis is active is based on the derivatives. The color of the node represents the largest magnitude of the derivative over all its CPT entries - see the node definition page to get the colors for the CPT elements.

The colors are scaled from white to red, with RGB(255, 0, 0) for the largest derivative in the network.

The gray nodes do not contribute anything, and are cut off by the relevance layer before actual derivative calculations are performed.

[PCherubini]

thank you!