Comparing Domain Authority Scores

[apo11o177]

Since your scale (like PageRank) is a logarithmic scale, it makes it hard to judge the distance between 2 scores.  Can you give me a rule of thumb.  For PageRank,  each jump is an exponential jump - so that a PR6 is perhaps 10 times stronger than a PR5.

What is the log base that SEOMoz uses.   Should I assume that a 60 is 10 times stronger than a 50?

This is important when it comes to measuring progress because growth is going to get more difficult as you move up the scale and I need to communicate the distance between our current Authority score and our goal.

Thank You!

[Dr-Pete]

Talked to Dr. Matt, and he said that, if you just want an estimate, take the log (base 10). You'll get a value from zero to two that will be roughly linear, and then you can scale it up to whatever range you need.

[Searchmint]

Hi Peter,

I have a follow up question to your reply; I want to try and use DA together with other linear metrics to create my own value that I can use for different purposes.

Considering your answer above it seems to me, and I have very little mathematical knowledge, that there is no exact log base I can use to get an "exact" representation of what a DA value is trying to express. Does that seem like a correct understanding?

Then, from a practical point of view, how would you recommend I convert DA values to a linear scale? Close enough is good enough i guess.

Thanks //David

[Dr-Pete]

I'm afraid the answer is "it's complicated" - here's the longer version, from Dr. Matt Peters, our head of Data Science:

"PA and DA are the output from a machine learning model that we then rescale to values between 1-100.  The raw output from the model is dimensionless and doesn't have any interesting meaning.  The rescaling is linear, but the inputs to the model are rescaled logarithmically before being used in the model.  We use the natural log (base e) but the base is pretty arbitrary since one can transform from one base to another by changing coefficients, and the coefficients themselves are set in a regression.  The key point is that since the inputs have a log applied to them it is much harder to increase DA from say 70 to 80 then it is from 30 to 40."