Thank you.
I tried to become familiar with page.rank and I guess it is an excellent idea but
I still have some problems. For instance, the results are highly variable
IGRAPH UNW- 328 11582 --
+ attr: name (v/c), x (v/n), y (v/n), weight (e/n)
First try:
page.rank (netz, personalized=c(2), damping=0.85)$vector[1:20]
rs72621144 rs76583332 rs116426014 rs9773882 rs71512836 156288
1.520751e-01 1.341455e-01 1.616638e-04 1.034574e-02 1.865783e-03 2.938891e-04
156375 rs9773471 rs7822515 rs7836416 rs7843227 rs11776856
3.365556e-04 1.061015e-02 3.606822e-04 1.026718e-02 1.062957e-02 1.666288e-03
rs10086982 rs10103818 rs7387067 rs78768549 rs117362856 rs28878202
3.298816e-04 7.856723e-05 1.059628e-02 1.051060e-02 1.066612e-05 5.524765e-05
rs13249796 rs3008288
1.250746e-02 1.411444e-03
Second try
page.rank (netz, personalized=2, damping=0.85)$vector[1:20]
rs72621144 rs76583332 rs116426014 rs9773882 rs71512836 156288
NaN NaN NaN NaN NaN NaN
156375 rs9773471 rs7822515 rs7836416 rs7843227 rs11776856
NaN NaN NaN NaN NaN NaN
rs10086982 rs10103818 rs7387067 rs78768549 rs117362856 rs28878202
NaN NaN NaN NaN NaN NaN
rs13249796 rs3008288
NaN NaN
Actually, I am wondering for the NaNs. I get highly variable resu Why are there NaNs for my vertex of interest (personalized=2).
A cutting of my adjacency matrix:
6 x 6 sparse Matrix of class "dgCMatrix"
rs72621144 rs76583332 rs116426014 rs9773882 rs71512836 156288
rs72621144 . 0.363491 . . . .
rs76583332 0.363491 . . 0.622703 0.219751 .
rs116426014 . . . . 0.259331 0.417368
rs9773882 . 0.622703 . . 0.219216 .
rs71512836 . 0.219751 0.259331 0.219216 . 0.366768
156288 . . 0.417368 . 0.366768 .
The results become less variable if I change the damping factor, for instance damping=0.5.
page.rank (netz, personalized=2, damping=0.5)$vector[1:10]
rs72621144 rs76583332 rs116426014 rs9773882 rs71512836 156288
5.022980e-01 2.525437e-01 5.145497e-05 5.698979e-03 1.567815e-03 7.693787e-05
156375 rs9773471 rs7822515 rs7836416
9.686269e-05 5.320426e-03 1.012695e-04 5.149335e-03
A damping factor closer to 0 (in comparison to the default of 0.85) makes it more likely to stay in the neighborhood of the personalised vertex. Is this correct? So a lower damping factor gives a better characterisation of the closely surrounding network. May I say that?
Do I get NaNs for damping=0.85 because the random walk ends far away of my vertex of interest?
If you are interested to check my igraph object, you are invited to download it via:
https://drive.google.com/file/d/0BxUQUYo5KHcLQ0UwNzd4R1BxdzA/view?usp=sharingThanks
Hermann