#16874: razib — 02/23  at  09:57 PM

Wait wait wait - “undergraduate degree from Caltech” == “non-trivial license to shoot one’s mouth off on all things sciencey”?

well, just like being a developmental biologist gives you creds to talk about networking dynamics.

#16875: Chris Clarke — 02/23  at  10:05 PM

well, just like being a developmental biologist gives you creds to talk about networking dynamics.

Being a trained forklift operator gives you sufficient cred to say that Kevin Drum has his head up his ass on this issue.

- Chris (Trained Forklift Operator)

#16879: yami — 02/23  at  10:43 PM
Yes, Razib, and normally I'd go along with your basic stance - this argument certainly isn't one where science-fu has played a very large role - but if there's this much special status in it for me and not just for our good host, I'm going to have to disagree. Sorry.

If my arguments seem half-baked, it's because you've misread my infallible determination of which situations do and do not merit full-fledged rigorous treatment. Bow to scientists! Bow to scientists now!

#16881: PZ Myers — 02/23  at  10:47 PM
I came out "blasting"? Here's my complete comment from Drum's site:

The ecosystem is managed by a male, using some semi-arbitrary algorithm he cooked up. It favors people who link up in lockstep -- check out some of those sites in the top tier. Most aren't particularly interesting or special, but they do engage in a lot of mutual linkage.

It favors conservatives. It particularly favors unthinking conservatives. It is an extremely poor measure of the diversity of the blogosphere, but does reward same ol', same ol', backslappin', good ol' boy networking.

I guarantee you that if a blog like What She Said! assembled a comparable "ecosystem" using their own algorithm, you'd find a consensus measure of popularity that would be heavily loaded with women and liberals. You're assuming that NZ Bear's biases are not influencing the rankings, and that's simply absurd.

What's the complaint? It's true: NZ Bear's algorithms are clearly simple and not very accurate, and produces a list using some kind of weighting that he picked. It's easily gamed. It is most definitely not an objective measure of relative quality or even popularity.

Anyone could invent a different algorithm that would produce a different list. For instance, design one with an inverse weighting scheme that discounts links from heavily trafficked weblogs; make a number of links from diverse smaller weblogs more valuable in ranking than a few links from those top tier weblogs. That would generate a differently ordered list than one that weights traffic from the big nodes more heavily, or one that doesn't try to weight links. And each model has a perfectly legitimate rationale behind it.

I don't even know why you people are arguing about this. It's obvious. Any time you try to compress a multidimensional measure into a single linear parameter, you've got a zillion different ways to do it and a zillion different ways to skew the results however you want...oh, wait, now I get it. Now I understand why the gnxp crowd objected to my comments. Reification of artifical, abstract simplifications of complexity is your thing. Criticizing arbitrary linear scales is bad for the IQ business.

And sorry, Amy, but you are completely wrong. I am not hold Kevin Drum to a different standard than I hold myself. We all need to work on this stuff; slapping 50% women and 50% men on my blogroll (or Drum's) would not make bias go away, and I have never claimed such a thing. What does help is putting mechanisms in place to think and review. No one should want to be handed a position because of their sex--what's right is that everyone should have an equal chance to be evaluated.

Locking the status quo in place and deluding yourself into thinking that there is an objective justification for it is not a way to correct the problem, and that's what Drum was doing. NZ Bear's "ecosystem" is a simplified construct loaded with implicit biases, not a ruler.

#16887: lloydletta — 02/23  at  11:25 PM
I hope I'm on your blogroll because my blog is decent, not because I'm female. I think there are more male blogs than female blogs which also skews the results.

I'd hate to see someone analyse my blogroll. I have more men than women on the blogroll.

It's pretty hilarious that people think my blog is written by a guy. Reminds me of the time that some idiots on the Bush2000 list accused me of being transgender.

#16889: — 02/23  at  11:48 PM
PZ - Just so there's no mistake, I don't frequent Gene Expression, and am a semi-regular reader here, though I haven't really felt strongly enough before to want to comment.

Also, as I wrote earlier, I don't have a problem at all with your later clarifications on this issue, but I felt I needed to disagree with my host, and I had no idea that I would get so under Rana's skin by doing so. My apologies.

My point stands- the ecosystem can be gamed by anyone. It isn't a matter of good old boy tactics, and I reject the charge that a female coded ecosystem would see women dominating the new system. You were simply way over the top, and I particularly found offensive that a women would purposely design a system that hi-lights women's issues and blogs, the implication being that women can't write about issues that appeal to a general audience. If women want to dominate the ecosystem they can do what the boys are doing, whatever works to game the system. Nothing male, or conservative about it. No male bias - that's a cop-out.

Lastly, I find it objectionable that you assume the worst about the coder of the ecosystem and the top blogs, namely that they have a gender bias, and when I read your post this morning and followed all of the links, I discover that your blogroll is no different. Simply the gene expression blog made a better argument in this case.

#16903: Dienekes — 02/24  at  01:55 AM
To determine whether author X is biased in favor of women or not, we need to do the following.

1) Measure the % of women who have blogs in X's subject area
2) Quantify the relative quality of female vs. male blogs (we may use the equality assumption if there is no easy way to do this)
3) Measure the % of blogs in X's blogroll, and compare it with (1) corrected by (2)

E.g., if X writes in an area where 25% of the blogger population is women, and he has 25% of women in his blogroll, then he is not biased: he just reflects the situation in his area. If he has 45% women and 55% men he is again not biased against women, he is actually biased in favor of women.

#16908: David Boxenhorn — 02/24  at  02:58 AM
PZ said (comment 53):

It’s true: NZ Bear’s algorithms are clearly simple and not very accurate, and produces a list using some kind of weighting that he picked.

NZ Bear doesn't use ANY weighting. You added this idea so that the contention that he inserted some secret nefarious bias would be plausible. Not very scientific, eh? More like a witch hunt, I would say. NZ Bear's algorithm is very simple and transparent, which is why he picked it: It's the number of unique inbound links.

He also ranks by traffic.

YOU are the one that wants to use some complicated weighting system in order to insert your own bias. Why don't you try it? I have a feeling that it won't achieve your objective.

#16933: PZ Myers — 02/24  at  08:10 AM
There's no such thing as no weighting. Of course he's weighting links: he's giving all of them equal weight. I'm also not proposing any complicated weighting system, and in particular, I'm not accusing anyone of "secret nefarious bias". As I plainly said above, any scheme that tries to put gnxp, pharyngula, and frogs&ravens on a linear scale is going to require assumptions and compromises. And it doesn't matter whether it's done by a conservative or liberal, they'll all have to diddle the numbers somehow.

He ranks by traffic modified by some odd stuff; you have to use sitemeter, which adds another condition to it. Sitemeter has its own weirdnesses. I've got my own network tools here and full access to my apache logs -- sitemeter undercounts inconsistently. You should know that logfile analysis isn't exactly a trivial thing, with lots of different programs out there, and the three I've used all produce different numbers from the very same logfile.

I'm not interested in putting together a ranking service. I wasn't interested enough to count heads in my own blogroll, remember? But you're quite wrong. If I put out a call for submissions to the Pharyngula Blogranking Protocol, I think it's obvious that the population of self-submitters here would be very different from the population if it were done at gnxp or at What She Said.

#16944: Chris Clarke — 02/24  at  09:18 AM
Technorati, at this writing, shows 59 sites with inbound links to my blog. TTLB Ecosystem puts that number at 27.
Interestingly enough, the most important link in my Ecosystem ranking is from this site, and I'm not in PZ's blogroll - it (and Technorati) picked up on the "Frequent commenters" link, meaning that I inadvertently gamed both systems to obtain a valuable link from a highly respected blog.

(Is that hagiographical enough, PZ? Amy seems to think Rana's set the bar pretty high.)

One fact that rather leaps out: TTLB includes not a single LiveJournal link to my blog, and there are at least a dozen. I don't know whether this is a fluke, but you need not postulate a deliberate conspiracy on NZ Bear's part to create a huge undercdounting of women: all you need to do is decide for some reason that LiveJournals are not "real blogs" or encounter some technical obstacle to combing them for links. (For whatever reason, LiveJournals are pretty heavily female.)

In any event, the discrepancy between my TTLB total and my Technorati total, indicates that there is something other than straight data collection going on. Clearly more study is needed, preferably done by someone who (unlike me) gives a flying fuck what some neocon says about my blog.

#16948: PZ Myers — 02/24  at  09:49 AM
Hey...you aren't on my blogroll?!?! That shouldn't be.

I took a look and found out why, though: I couldn't find a syndication link. A while back, I reorganized everything to use my newsreader exclusively to maintain the list of blogs I read (it's the most efficient way to scan a long list, and I really do at least skim every blog on the blogroll), which meant you dropped out. If you've got an RSS/XML/RDF/Atom link, send it and you'll be on there right away.

That's another example of unintended biases, by the way. It's also hard to find feeds on LiveJournal sites and on some sites that use non-MT/Typepad/blogspot software.

#16950: Chris Clarke — 02/24  at  09:53 AM
My rdf file is here. Thanks, PZ.

#16951: Chris Clarke — 02/24  at  09:53 AM
My rdf file is here. Thanks, PZ.

#16953: Chris Clarke — 02/24  at  09:54 AM
Sorry for the posting glitch. First link has a typo. Thought I caught it in time to avoid the post.

#16982: David Boxenhorn — 02/24  at  11:36 AM
PZ (comment 58):

He ranks by traffic modified by some odd stuff; you have to use sitemeter, which adds another condition to it. Sitemeter has its own weirdnesses.

There you go again with your insinuations! NZ Bear uses straight Sitemeter readings. It's true that Sitemeter has its own method of counting visits (since the World Wide Web is stateless, you have no choice but to invent your own method) but it is completely transparent and there is no reason to think it is male biased. Why is your unsubstantiated default assumption that it is male biased? And it seems that it has never occurred to you that NZ Bear used it for no other reason than that it is available!

Your default assumption that he is guilty of bias because he is male, itself reeks of bias. I might as well assume that same about you.

#16985: PZ Myers — 02/24  at  11:45 AM
Where did I say sitemeter had a male bias? I don't know what biases using sitemeter would incorporate. This is not an insinuation: it's a straight up fact. Using self-selection, using a particular restricted toolset, calculating values in particular ways...all impose biases on an outcome. And it certainly did occur to me that Bear selected it because it was available--that was my assumption. Availability is another factor that imposes a bias. And yes, being male biases things for all of us males, as does being female.

The problem here is obvious. You are reading too much into a word. I say "bias", and I'm thinking about unavoidable factors that influence results. You see me use the word, and you're thinking "Ah, he believes in nefarious, devious schemes calculated to cheat!"

Get over it. Any weird freight like that added to what I've said is your invention, not mine.

#17006: David Boxenhorn — 02/24  at  12:55 PM
Where did I say sitemeter had a male bias?

Well, Mr. Myers, if you were not talking about male bias, then why are you changing the subject? No one here is interested in discussing random bias.

#17010: Rana — 02/24  at  12:59 PM
Goodness. So now defending a person against unfair attacks is deifying them? Interesting.

fwiw, AmyJ, I spent this much time here largely because I had a dull day at work, and because it is entertaining to me to force people to reveal their underlying axes that they try to grind. I am also amused by people who accuse others of hypocrisy, bad arguments, inability to understand the situation, and lack of focus while demonstrating those same traits themselves.

You keep accusing me of "flitting" and of misunderstanding things, and so on. Yet you yourself try to make the debate into something it's not (including misreading what PZ meant by "Fan Club" -- hint, he was being sarcastic, and it involved the Gene Expression folks), fail to respond to my challenges on that count, and keep trying to dodge my main point -- it isn't about the blogroll count per se; it's about using inappropriate data to accuse of PZ of hypocrisy while ignoring hypocrisy on the part of the person he challenged. So my posts were (a) establishing context, (b) pointing out the inappropriateness of razib's data, (c) tossing in some additional questions about that data's accuracy for the hell of it, (d) noting that this is not simply about PZ's blogroll but about the larger question of how a blogger demonstrates feminist sympathies, (e) and finally, questioning a person who _claims_ that she's only interested in rooting out hypocrisy -- in a situation where hypocrisy is _not_ present, and while ignoring a situation where it _is_ present.

Shorter version, since you seem to need one (I thought you were smarter than this -- sorry):

Razib accused PZ of hypocrisy, using a standard that PZ himself called into question. His data was not relevant. There are other standards by which to judge whether PZ lives up to espoused feminist values. PZ meets those standards. He is therefore not a hypocrite.

Kevin Drum, who reopened a known can of worms, identified a problem that could be solved by linking to more women. He claimed to be worried about this problem, suggesting that he'd like it to be solved. He then fails (and has failed) to act on his own suggestion, even though it would be a simple thing to do. I'd argue that this makes _him_ a hypocrite, though I'd be willing to yield some ground on this.

Even shorter version: Kevin was probably hypocritical. PZ, who critiqued him, was not. Attempts to apply Kevin's standards in order to prove that PZ's a hypocrite are interesting but not relevant. Attempts to brand PZ a hypocrite while ignoring Kevin's probable hypocrisy are also interesting, but also irrelevant, except in as how they reveal the biases of the accuser.


Is that clear and simple enough for you?


On a related issue: yes, the Ecosystem does not count links and hits reliably. I average around 200 hits a day, according to Sitemeter (which TLB claims to draw from, yet none of that information shows up in the ecosystem's description of my blog, something I've never understood). I know for a fact that I have at least a hundred people linking to me (because they are all on my blogroll; Technorati would claim that I have 218 inbound links from 111 sources, but given that that number never changes, I have doubts as to its accuracy), yet the Ecosystem has my links and linking blogs at about 47.

The Ecosystem may well work for a certain subset of blogs, but it's not reliable across the board. Thus any claims about blogs in general cannot be effectively defended by recourse to the Ecosystem alone.

#17023: David Boxenhorn — 02/24  at  02:59 PM
For the record (http://www.truthlaidbear.com/EcoFAQ.php):

What links does the TTLB Ecosystem count?

All links from a scanned weblog to any other weblog (i.e., links from a weblog to itself are ignored). This includes links within posts, and 'permanent' links in a weblog's blogroll.

Does the TTLB Ecosystem scan all pages of a weblog?

No, only the front page.

Does the TTLB Ecosystem handle Blogrolling.com blogrolls?

Yes. Even if you are using the Javascript implementation of Blogrolling.com, the Ecosystem is designed to retrieve your blogroll when it scans your weblog.

And if you want to participate in the Sitemeter rankings, you have to make your stats public.

There's nothing mysterious about it. (Technorati is another matter...)

#17037: Chris Clarke — 02/24  at  05:58 PM
It was particularly interesting to me to see how AmyJ interspersed her argument - much of which was well-written and persuasive – with particularly gender-loaded criticism of Rana. "fawning..." "losing the big picture..." "you’re like a hummingbird, zipping all over the place. Please stop spinning in circles."

Contrast this with her beginning comment to PZ, who was her putative target of criticism:

"PZ, you're witty as always."

I'm old enough and have known enough women to not be surprised that some women reserve their real venom for other women, while remaining cordial and professional with men even when they disagree strongly. Still, it was amusing to me to see a new variation on the sixth contributor to this hypothetical blog post.

#17052: jinnderella — 02/24  at  11:02 PM
While it is silly to compare gnxp to PZ's blog (since one is a group blog while the other is sole proprietor), i do have some statistics.
From my MBTI poll last month, gnxp has a solid 26% XX commentor/readership. Of the total 72 teammembers with posting accounts 8 are XX, or 1/9. Of the last 15 entries, including the tributary scifi blog, three are by XX authors. Given the proclivity of my sex to be mouthy, I think this is accurate.

Chris Clarke: You speak truth. On a recent visit to Drum's site I was nic-jacked, told i must be attractive to be so well treated by the XY, and called the "Uncle Tom of the Women" by "Nancy".

#17198: Rana — 02/26  at  03:52 PM
David -- thanks for posting that quote. I do have some questions, still, especially about this part: "What links does the TTLB Ecosystem count? All links from a scanned weblog to any other weblog..."

This is odd, because it implies that ranking is determined not by who links TO a blog, but by how many links _originate_ from the ranked blog. If so, it's not an indicator of popularity so much as of one's fondness for outlinking.

Even if we interpret this as a statement that ranking is determined by iincoming links, there's a problem with that assertion that links from _any_ other weblog are counted. This, in my own case, is patently false. What it probably should say is "All links to a scanned weblog from any other _scanned_ weblog." That is, the Ecosystem does list blogs that link to me that are already in the Ecosyste (but, again, not all of them) and it doesn't notice many that I know for a fact have linked to me (because I've seen their blogrolls).

Moreover, it is inconsistent in counting the number of links I get from any particular blog. If a blog links to me twice, shouldn't both links count? If not, shouldn't it then be one link per blog for all blogs? But that's not what's going on. Some blogs count only as one link, even though they've linked to me more than once, while others' multiple links _are_ all counted.

All of these discrepancies make me less inclined to trust the Ecosystem as a yardstick for blog popularity in light of that quoted explanation , not more.

(It may be, in part, to a problem handling blogroll links vs. in-post links. Maybe.)

#17199: Rana — 02/26  at  03:55 PM
Chris, I noticed that gendered rhetorical dynamic too. Believe you me, I was trying very hard not to use gendered dismissive language in my own responses to her. (It was tempting, I admit!)

#18390: Chris Clarke — 03/11  at  04:48 PM
This just in (via Sean): ranking blogs based on the number of incoming links may in fact produce a ranking that favors conservative blogs.

#18392: PZ Myers — 03/11  at  05:10 PM
I notice that they specifically show that there are different ranking schemes that generate different arrangements of the top-ranked weblogs.

I am vindicated.