4 ms·

With smaller data sets I find the summary to be less valuable than the actual paper. In this case we are talking about 4,533 twin pairs where one is a male and

by belorn 8y ago

With smaller data sets I find the summary to be less valuable than the actual paper. In this case we are talking about 4,533 twin pairs where one is a male and the other female. The female twin had a 2.8-percentage point higher probability of dropping out of high school compared to other female students, and a similar 1.9-percentage point lower probability of graduating from college. Those two numbers seems connected, ie that individuals that have higher risk of high school drop out have also a higher risk of not graduating. The study do not cross compare this with male students which also have higher drop out rate and lower graduation rate compared to female students, but some relation seems implied. The study also notes a lower probability of ever having been married by age 32, and lower probability of employment. The article does not explore if the effect could be exclusively a result of lower education. As a final thought, the sample size is relatively small and the effect relatively minor in absolute numbers. The paper does not give out any P values or confidence intervals for their primary findings. The relative effect is large, but I interpret that more like a warning flag. In absolute terms we are talking about 136 additional students born between 1967 and 1978 that dropped out compared to a null hypothesis. The same number for college drop out was 49 students that did not graduate. With so few individuals having that large effect on the conclusion I personally will wait a bit for replication studies.

tofof 8y ago
> The paper do not give out any P values [...] for their primary findings. The relative effect is large, but I interpret that more like a warning flag. Incorrect. The p-values -- all 85 of them -- are presented in the appendix[1], linked to in the first paragraph under the 'Results' heading, along with means, standard deviations, and number of observations. The article text[2] repeatedly references these tables when making each claim. > The paper do not give out any [...] confidence intervals for their primary findings. Incorrect. Confidence intervals are presented in 100% of the figures. > The article does not explore if the effect [on marriage or employment] could be exclusively a result of lower education. Incorrect. They explicitly consider the effect of educational attainment on future economic earnings. "Behavioral changes downstream of prenatal testosterone exposure, such as any effect of disruptive behavior on schooling or educational attainment, are likely candidate pathways linking male co-twin exposure to later earnings. Consistent with this interpretation, conditioning on educational attainment in the earnings equation reduced the coefficient on opposite-sex from −0.086 (SE, 0.021) to −0.042 (SE, 0.014), suggesting that roughly 50% of intensive margin labor market effects could be explained by educational outcomes that are upstream of earnings. > Those two numbers [H.s. dropout rate / college graduation rate] seems connected ... some relation seems implied. These measures are already known to be strongly correlated, along with the correlations to the probability of inclusion within the labor force and lifetime earning potential. The paper makes no claims that these four measures are independent from one another. The paper does explicitly make clear when findings of independence are relevant, particularly in usually-correlated variables. In this case I'm referring to their findings of a fertility rate effect independent from marriage rate. 1: https://www.pnas.org/content/pnas/suppl/2019/03/14/1812786116.DCSupplemental/pnas.1812786116.sapp.pdf https://www.pnas.org/content/pnas/suppl/2019/03/14/181278611... 2: https://www.pnas.org/content/early/2019/03/14/1812786116 https://www.pnas.org/content/early/2019/03/14/1812786116
- belorn 8y ago
  > are presented in the appendix[1], Thank you, missed the appendix. Also see that in the figure page list CI as 95%. > These measures are already known to be strongly correlated... The paper does explicitly make clear when findings of independence are relevant The first few paragraphs in the article here and the Significance section in the actually paper imply a strong independence. My point above is basically that the independence looks to be much weaker than a casual read might imply. It also boast with the largest study ever done on the subject, which might be true, but the absolute number are actually quite low with only a few individuals being responsible for a large effect in the data. That is one of the biggest warning signs that exist in this kind of research. I thus get skeptical, at which point I start to look why there seems to be a disconnect.
thomasahle 8y ago
Does it compare them to girl-girl twins?
- tofof 8y ago
  Sort of. It compares female-female to male-female twin pairs in analyzing subsequent maternal fertility, to ensure the relevance of the deceased co-twin approach. Search for 'Table S5' in the paper and the appendix.

by belorn 8y ago

tofof 8y ago

> The paper do not give out any P values [...] for their primary findings. The relative effect is large, but I interpret that more like a warning flag. Incorrect. The p-values -- all 85 of them -- are presented in the appendix[1], linked to in the first paragraph under the 'Results' heading, along with means, standard deviations, and number of observations. The article text[2] repeatedly references these tables when making each claim. > The paper do not give out any [...] confidence intervals for their primary findings. Incorrect. Confidence intervals are presented in 100% of the figures. > The article does not explore if the effect [on marriage or employment] could be exclusively a result of lower education. Incorrect. They explicitly consider the effect of educational attainment on future economic earnings. "Behavioral changes downstream of prenatal testosterone exposure, such as any effect of disruptive behavior on schooling or educational attainment, are likely candidate pathways linking male co-twin exposure to later earnings. Consistent with this interpretation, conditioning on educational attainment in the earnings equation reduced the coefficient on opposite-sex from −0.086 (SE, 0.021) to −0.042 (SE, 0.014), suggesting that roughly 50% of intensive margin labor market effects could be explained by educational outcomes that are upstream of earnings. > Those two numbers [H.s. dropout rate / college graduation rate] seems connected ... some relation seems implied. These measures are already known to be strongly correlated, along with the correlations to the probability of inclusion within the labor force and lifetime earning potential. The paper makes no claims that these four measures are independent from one another. The paper does explicitly make clear when findings of independence are relevant, particularly in usually-correlated variables. In this case I'm referring to their findings of a fertility rate effect independent from marriage rate. 1: https://www.pnas.org/content/pnas/suppl/2019/03/14/1812786116.DCSupplemental/pnas.1812786116.sapp.pdf https://www.pnas.org/content/pnas/suppl/2019/03/14/181278611... 2: https://www.pnas.org/content/early/2019/03/14/1812786116 https://www.pnas.org/content/early/2019/03/14/1812786116

belorn 8y ago
> are presented in the appendix[1], Thank you, missed the appendix. Also see that in the figure page list CI as 95%. > These measures are already known to be strongly correlated... The paper does explicitly make clear when findings of independence are relevant The first few paragraphs in the article here and the Significance section in the actually paper imply a strong independence. My point above is basically that the independence looks to be much weaker than a casual read might imply. It also boast with the largest study ever done on the subject, which might be true, but the absolute number are actually quite low with only a few individuals being responsible for a large effect in the data. That is one of the biggest warning signs that exist in this kind of research. I thus get skeptical, at which point I start to look why there seems to be a disconnect.

thomasahle 8y ago

Does it compare them to girl-girl twins?

tofof 8y ago
Sort of. It compares female-female to male-female twin pairs in analyzing subsequent maternal fertility, to ensure the relevance of the deceased co-twin approach. Search for 'Table S5' in the paper and the appendix.