luminosis.blogspot.com is probably written by a male somewhere between 66-100 years old. The writing style is personal and happy most of the time.
There isn't a heck of a lot of detail about how the evaluation is done, but presumably it is based on word usage analysis. Based on the words I use, this algorithm thinks I am mostly happy, but possibly because of the formality of my sentence construction, thinks I am between 66 to 100. Well, it got the gender right.
Turning this lens on some other examples -- Michael and Jill's Second Draft, as well as Madeline's Escaping The Trunk gets exactly the same evaluation as me, except for the gender -- the algorithm correctly identifies Madeline as female, and believes that the combination of Michael and Jill is also female. Maybe because Jill posts more? The analysis is dead on for Dave -- male, 26 to 50, mostly happy. But for Peter, it gets both the gender and age wrong -- female, 66 to 100, mostly upset. Looking further afield to the blog of a complete stranger -- the age problem and "mostly upset" rating also applies to Pharyngula. Maybe the "mostly upset" part is a marine biologist thing?
These results are dynamic over time, since the analysis is based on the latest articles in the blog. It is possible that I could change the tone and character of my blog and get a different result, so if you are reading this article years after it is written, and you run the same analysis on my blog, you might not get the same result.
You likely can derive information about an author from lexical analysis of their writing, including statistical frequency analysis of their word usage. But these (admittedly limited) results show this analysis needs to used in context with other information. In the meantime, I need to go find some kids so I can yell at them to get off my damn lawn. Because that's what us happy old fuddy duddy bloggers do.