search

thongphamthe / PAFit

PAFit source
https://CRAN.R-project.org/package=PAFit
17 stars 1 forks source link

Changing s-parameter in PAFit #4

Open eglantine-coder opened 3 years ago

eglantine-coder commented 3 years ago

Hi,

I've been getting very high s-values when running the PAFit model on my data. How can I decide the value, especially when wanting to compare results for different datasets?

I've tried changing the parameters within the PAfit function but it doesn't work.

Thanks

thongphamthe commented 3 years ago

Hi, A high value of s is perfectly fine. It means that node fitnesses are highly concentrated. For interpretations, instead of s, it may be better to look at the estimated attachment exponent and the estimated fitness distribution.

eglantine-coder commented 3 years ago

Hi,

Thank you for the quick reply.

I have attached two examples of the results I got for the PAFit model on two different datasets. How could I interpret the 1.06 value for United States vs. the 53.8 for UAE in terms of PA and fitness ? Are they comparable at all ?

Many thanks

results results
thongphamthe commented 3 years ago

Hi, Fitnesses (and PA) from different datasets are not comparable.

eglantine-coder commented 3 years ago

Hi,

Then, how could we compare the estimated attachment exponent and the estimated fitness distribution, given the following results, for instance?

stats stats

Thanks

thongphamthe commented 3 years ago

The strengths of the PA effect and the fit-get-richer effect can be compared between different datasets.

In your example, one can compare the estimated attachment exponents and say that the PA effect in the second dataset is stronger than the PA effect in the first dataset.

For comparing the fit-get-richer effect, one way is to compare the variances of the estimated fitness distributions. A larger variance implies a stronger fit-get-richer effect.