You're looking at Less Wrong's discussion board. This includes all posts, including those that haven't been promoted to the front page yet. For more information, see About Less Wrong.

Viliam comments on Open Thread, May 4 - May 10, 2015 - Less Wrong Discussion

3 Post author: Gondolinian 04 May 2015 12:06AM

You are viewing a comment permalink. View the original post to see all comments and the full post content.

Article Navigation

Comments (215)
Tags: open_thread

Comments (215)

You are viewing a single comment's thread. Show more comments above.

Comment author: SanguineEmpiricist 04 May 2015 03:14:08AM *  3 points [-]

Interested in what you guys think about this. Jayman(hbd blogger) say's parenting has no effect on how children turn out. Seems empirically incorrect to me and it's just probably difficult to encapsulate the results/hard to see non-linearities to make it easy to reference.

He insists on twin-adoption studies contrary to my views.

Thoughts? This sort of seems like the two cultures divide we agree on. I might make a thread just for this.

Argument: Does parenting have any effect on child outcomes?

  • His view: Zero effect & Breast milk confers no advantage either

  • My view: Parenting has some variable effect that is difficult to encapsulate in the studies he references while maintaining the correctness and good taste of genetic arguments.

I cite decision theory, statistical inference, study design, and the related area as being primal over empirical references which have failed to encapsulate the effects he is pointing to in his observables. Statistical inference just doesn't work like that to give such strong conclusions. Any one who reads the literature on study design/inference knows that it's just not possible to give recommendations that are that strong. Sort of in the realm is Isaac Levi's "Gambling with the Truth" if not only the first few chapters although not quite, probably just statistical study design/inference in general.

Thoughts any one? I think scientists or empirical researchers are not used to being told that there is a higher plane of reference. Saying that there is zero influence is equivalent to saying all the relevant variables have been enumerated and assigned exact values for probability & effect and that there is nothing else to be assigned.

I believe my orientation is correct.

edit: I might add that not ONLY would that be saying that the relevant variables have been completely enumerated && assigned cost functions but that we are sure there is nothing else(no uncertainty) and that we are sure they all equal zero/canceled out.

Comment author: Viliam 06 May 2015 07:37:35AM 5 points [-]

His view: Zero effect & Breast milk confers no advantage either

Breast milk "confers no advantage" compared to what?
(baby dying from starvation? water? cow milk? soylent? infant formula -- which one?)

All that stuff that exists in breast milk -- such as immunoglobulin A, lysozyme, blood albumin, creatine -- it's just completely useless, I guess. I wonder why evolution even bothered to design such system, when it provides no advantage. (This is a sarcastic way to say "my priors for this hypothesis are very low".)

Comment author: SanguineEmpiricist 06 May 2015 07:49:03AM *  2 points [-]

Wait did you mean for Jayman's hypothesis is low or breast milk?

Comment author: gjm 06 May 2015 10:24:23AM 3 points [-]

(I'm not Viliam, but:) For Jayman's hypothesis, obviously. The argument is: Here's all this complex machinery put in place by evolution; it's terribly unlikely that what it does is actually useless.

(That would argue that breast milk is a good food for babies, which I don't think anyone denies. It's only a strong argument against feeding them infant formula in so far as we have reason to think that formula doesn't have in it those things that evolution helpfully put into human breast milk.)

Comment author: Viliam 06 May 2015 03:45:40PM *  2 points [-]

Yes, low priors for Jayman's hypothesis. Although it wasn't sufficiently specified what exactly his hypothesis was, so I could be arguing against a strawman.

I am not an expert on infant formulas, but I think that if someone could factory-produce a drink that updates your immune system (as breast milk does), that would have huge implications in medicine. Essentially, we could replace vaccination by drinking soylent.

Comment author: Douglas_Knight 06 May 2015 11:54:00PM -1 points [-]

There have been many randomized controlled experiments of breast milk vs formula. Every single one of them has shown no effect.

Comment author: gjm 07 May 2015 01:11:34AM 4 points [-]

I just did a little googling and that appears to be untrue. What's the source of your information?

The first hits from a Google search for <<random controlled trial breast formula milk>> were:

  • Scholarly articles (Google pulls out a few of these at the top of its search results):
    • A study on promotion of breastfeeding in Belarus. Intervention appears to have increased breastfeeding and also substantially (and significantly) reduced the incidence of two of the three health problems they measured.
    • A study on how often breastfeeding passes on HIV from mother to child. (Conclusion: about 1/6 of the time; using formula is an effective way of reducing this. Interesting but not relevant here.)
    • A study on the effect of "peer support" on making mothers breast-feed for longer. Not relevant here.
  • Ordinary search results:
    • First five hits are for the same study, looking at very premature babies. Conclusion was that mother's milk appears better than both donor breast milk and formula, and the latter two aren't much different.
    • Meta-analysis of studies comparing donor breast milk with formula. Conclusion is that unfortified donor breast milk (i.e., no extra nutrients added) leads to slower growth than formula, but formula leads to more necrotizing enterocolitis. Apparently donor breast milk is generally fortified nowadays.
    • Different meta-analysis looking specifically at necrotizing enterocolitis. Found only four relevant trials, from 20 years before. They were too low-powered to find significant results, but aggregating them finds a substantial and significant reduction in NEC risk for donor breast milk versus formula.
    • CBS News report on a study that found that giving a little formula to babies who lose a lot of weight shortly after birth may result in longer breast-feeding. Interesting but not obviously relevant.

And that's the end of the first page. So it looks to me as if

  • There haven't been that many randomized controlled experiments -- fully half the first page of Google results were from a single one, and the meta-analyses seem to have found rather few, rather old, rather small studies.
  • Some small old studies didn't find significant results.
    • But that seems to have been because of lack of power rather than because differences weren't there.
    • Aggregating those studies finds significant differences.
  • There may be relevant differences between mother's milk and donor milk; comparing mother's milk to formula in RCTs is probably harder than comparing donor milk to formula. (Because of ethical difficulties in telling mothers not to breast-feed their children; because of practical difficulties telling mothers to breast-feed their children if they have trouble with that; because of difficulties in blinding. Maybe other reasons.)

Which seems like pretty much the reverse of what you say. But of course looking briefly at one page of Google results is not a proper scholarly literature review; would you like to tell us more?

Comment author: SanguineEmpiricist 07 May 2015 02:37:05AM 3 points [-]

http://www.bayesianrisk.com/chapters.html http://www.bayesianrisk.com/sample_chapters/Chapter%201%20There%20is%20more%20to%20assessing%20risk%20than%20statistics.pdf

I agree with Viliam & gym. This just points to the limits of statistical knowledge & that we need to supplement with other logical-experimental knowledge, such as arguments from evolution.

Risk cannot only be based on statistical knowledge, as chapter one of bayesian risk argues.

Comment author: TheAncientGeek 06 May 2015 09:38:07AM 0 points [-]

compared to what? 

Always a good questions to ask.

Powered by Reddit Powered by Reddit