John Fuerst
Dalliard
Abstract
The authors conducted a meta-analysis of behavioral genetic variance components (ACE) x race/ethnicity interactions for cognitive ability. The differences between the variance components for Black and White Americans were small, despite the large average test score differences. More substantial differences were found between Hispanics and Whites, though results were based on only two studies.
Attached:
Doc File
Excel file
I fixed the title. -Emil
Back to [Archive] Post-review discussions
Excellent paper. I have some comments. Numbers refer to lines in the file.
24:
"Such analyses directly indicate only the sources of within-race differences, but they nevertheless have important implications for understanding the causes of the Black-White and Hispanic-White mean differences in IQ, which research consistently shows to be about 1 and 0.7 standard deviations, respectively (Roth et al., 2001). "
I would also add this meta-analysis: http://humanvarieties.org/2013/01/15/secular-changes-in-the-black-white-cognitive-ability-gap/
It is much newer (12 years) than the Roth one. Note also that some people have recently claimed that the W-B gap is diminishing. It was already discussed in 2005 in the special issue about Rushton and Jensen's review paper. Perhaps you want to mention this. Perhaps not.
Space too much at the end.
28:
"In behavioral genetics, the sources of IQ variance can be partitioned into three components: heredity (h2), shared environment (c2), and unshared environment (e2). These are also known as the ACE components. The meaning of h2 (also referred to as a2) should be obvious. c2 refers to environmental effects that serve to make family members more similar to each other, while e2 consists of those non-genetic effects that are not shared between family members but differentiate them from each other. c2 and e2 are collectively known as environmentality. The basic biometric model assumes that environmental and genetic influences are additive, but there may also be interactions between them, and these can be estimated as well."
I would include a source for these claims. For instance, Plomin et al 2012. https://www.goodreads.com/book/show/15870785-behavioral-genetics
Jensen's publications also introduce the concepts well.
51:
"for lower social class groups, at least in the U.S. (Turkheimer and Horn, 2014). "
Space at the end.
53:
"narrative reports have drawn opposing conclusions (cf. Jensen, 1998; Scarr, 1981). "
Spacing.
55:
"We conducted a literature search for papers containing either heritability estimates or kinship correlations that would allow for the computation of such estimates for racial and ethnic groups in the U.S."
You need to provide explicit search criteria. Which databases did you search? Which keywords? When? Any limits on publication year?
Also explicit exclusion criteria.
57:
"ethnic groups in the U.S. When"
Spacing.
59:
"estimates. We were able to locate thirteen studies; these are shown in table 1 below. "
Spacing.
60:
"Table 1: List of Samples Included and Excluded"
I would bold the captions of the tables.
63:
"excluded. One study"
Spacing.
65:
"Beaver et al. (2013)"
Spacing.
69:
"excluded. This"
Spacing.
71:
"below. "
Spacing.
There are lots of more spacing errors. Just do a search for " " (two spaces, forum software overrides my text) in the document to find them.
81:
"(personal communication, October, 3, 2013; personal communication, September, 24, 2013) "
Don't need to repeat "personal communication", you can repeat the dates if you want or not.
202:
The table needs a caption and needs to be shrunk to fit the page. I would also like to have a column with sample sizes as this helps the interpretation.
280:
"Tucker-Drob, E. M., Briley, D. A., & Harden, K. P. (2013). Genetic and environmental influences on cognition across development and context. Current Directions in Psychological Science, 22, 349-355."
They stress the SES x h^2 correlation. They don't mention that other, very large, studies have not found the same result. Mentioned here: http://drjamesthompson.blogspot.dk/2014/01/ses-and-heritability-of-intelligence.html
24:
"Such analyses directly indicate only the sources of within-race differences, but they nevertheless have important implications for understanding the causes of the Black-White and Hispanic-White mean differences in IQ, which research consistently shows to be about 1 and 0.7 standard deviations, respectively (Roth et al., 2001). "
I would also add this meta-analysis: http://humanvarieties.org/2013/01/15/secular-changes-in-the-black-white-cognitive-ability-gap/
It is much newer (12 years) than the Roth one. Note also that some people have recently claimed that the W-B gap is diminishing. It was already discussed in 2005 in the special issue about Rushton and Jensen's review paper. Perhaps you want to mention this. Perhaps not.
Space too much at the end.
28:
"In behavioral genetics, the sources of IQ variance can be partitioned into three components: heredity (h2), shared environment (c2), and unshared environment (e2). These are also known as the ACE components. The meaning of h2 (also referred to as a2) should be obvious. c2 refers to environmental effects that serve to make family members more similar to each other, while e2 consists of those non-genetic effects that are not shared between family members but differentiate them from each other. c2 and e2 are collectively known as environmentality. The basic biometric model assumes that environmental and genetic influences are additive, but there may also be interactions between them, and these can be estimated as well."
I would include a source for these claims. For instance, Plomin et al 2012. https://www.goodreads.com/book/show/15870785-behavioral-genetics
Jensen's publications also introduce the concepts well.
51:
"for lower social class groups, at least in the U.S. (Turkheimer and Horn, 2014). "
Space at the end.
53:
"narrative reports have drawn opposing conclusions (cf. Jensen, 1998; Scarr, 1981). "
Spacing.
55:
"We conducted a literature search for papers containing either heritability estimates or kinship correlations that would allow for the computation of such estimates for racial and ethnic groups in the U.S."
You need to provide explicit search criteria. Which databases did you search? Which keywords? When? Any limits on publication year?
Also explicit exclusion criteria.
57:
"ethnic groups in the U.S. When"
Spacing.
59:
"estimates. We were able to locate thirteen studies; these are shown in table 1 below. "
Spacing.
60:
"Table 1: List of Samples Included and Excluded"
I would bold the captions of the tables.
63:
"excluded. One study"
Spacing.
65:
"Beaver et al. (2013)"
Spacing.
69:
"excluded. This"
Spacing.
71:
"below. "
Spacing.
There are lots of more spacing errors. Just do a search for " " (two spaces, forum software overrides my text) in the document to find them.
81:
"(personal communication, October, 3, 2013; personal communication, September, 24, 2013) "
Don't need to repeat "personal communication", you can repeat the dates if you want or not.
202:
The table needs a caption and needs to be shrunk to fit the page. I would also like to have a column with sample sizes as this helps the interpretation.
280:
"Tucker-Drob, E. M., Briley, D. A., & Harden, K. P. (2013). Genetic and environmental influences on cognition across development and context. Current Directions in Psychological Science, 22, 349-355."
They stress the SES x h^2 correlation. They don't mention that other, very large, studies have not found the same result. Mentioned here: http://drjamesthompson.blogspot.dk/2014/01/ses-and-heritability-of-intelligence.html
This version has multiple problems. I didn't know that John has already posted it here. So, I have responded by mails. Perhaps John will reply it directly ?
I am currently reworking the paper to meet Emil and MH's requests.
I am currently reworking the paper to meet Emil and MH's requests.
Ok, here it my latest draft. I addressed Emil's concern and included a discussion of search methods.
sp = spacing problem
6:
sp
7:
Dalliard's name is misspelled.
18:22
Needs a cite for claims. Pick one of the recent review articles. Last sentence should probably get "in adults".
32:
You mean "genetic" (broad). E includes error, which is worth mentioning because authors rarely correct for test unreliability so their results are underestimates as pointed out by Jensen in his '69 paper.
33:
h2 isnt the same as a2. a2 refers to additive genetic factors. h2 may or may not include more depending on whether it is broad or narrow. Narrow h2 is the same as a2.
40:46
This overlaps with the beginning of the Introduction. Rewrite.
56:
sp
57:
Odd phrasing. “Meta-analytic method” maybe? Or just “Method”?
60:
sp. There are so many more of these. Just search for “[space][space]” in the text editor to find them.
83:84
Rewords. Desegregated sounds like you are saying they live in racially segregated communities as in Apartheid times.
Tables have visual problems.
100:
Odd phrasing. “Results” or “Meta-analytic results” or something.
169:
I don't understand the results in Table 7. There are clearly non-zero values in Unstandardized col. E, but they are exactly 0 in Standardized. Explain?
283:
Where is “the first table”? I can't find a table with results for the g factor.
Can you conduct MCV analyses to see whether the g-loading of the subtest correlates with the h2 of it as claimed by among others Jensen (1998)? I note that the A for the g factor is higher than that reported in the Table 8, which indicates that it is the g factor that drives the h2, not the remaining variance.
Added. From looking the later tables, there are only 5 subtests with low heterogeneity, so MCV is probably not worth doing.
291:
Both Hart et al studies (2010, 2013) are missing from the reference list. You should check for more missing references.
It seems to me that research based on extended relations would be theoretically superior, not less robust. I could not check the Hart reference to see if some justification was given in that for this claim. Explain?
303:
You should conduct a test to see if this difference is reliable. Fluke perhaps? It goes against all standard theory that h2 should be lower for whites.
338:
Actually, the three way pattern of results is exactly reverse. In sibs: W>H>B. In extended, B>H>W. Very strange. Analytic error? I note that it is found using both methods of dealing with age from looking at Table 12.
353:
Sampling error does not seem a plausible candidate. All sample sizes are in the thousands. From looking at Table 12, the smallest sample is N=1112. Very large for a BG study.
367:
The extended model, does that also use siblings, or only cousin and further away data? Perhaps try estimates based only on e.g. cousins.
6:
sp
7:
Dalliard's name is misspelled.
18:22
Needs a cite for claims. Pick one of the recent review articles. Last sentence should probably get "in adults".
32:
You mean "genetic" (broad). E includes error, which is worth mentioning because authors rarely correct for test unreliability so their results are underestimates as pointed out by Jensen in his '69 paper.
33:
h2 isnt the same as a2. a2 refers to additive genetic factors. h2 may or may not include more depending on whether it is broad or narrow. Narrow h2 is the same as a2.
40:46
This overlaps with the beginning of the Introduction. Rewrite.
56:
sp
57:
Odd phrasing. “Meta-analytic method” maybe? Or just “Method”?
60:
sp. There are so many more of these. Just search for “[space][space]” in the text editor to find them.
83:84
Rewords. Desegregated sounds like you are saying they live in racially segregated communities as in Apartheid times.
Tables have visual problems.
100:
Odd phrasing. “Results” or “Meta-analytic results” or something.
169:
I don't understand the results in Table 7. There are clearly non-zero values in Unstandardized col. E, but they are exactly 0 in Standardized. Explain?
283:
Where is “the first table”? I can't find a table with results for the g factor.
Can you conduct MCV analyses to see whether the g-loading of the subtest correlates with the h2 of it as claimed by among others Jensen (1998)? I note that the A for the g factor is higher than that reported in the Table 8, which indicates that it is the g factor that drives the h2, not the remaining variance.
Added. From looking the later tables, there are only 5 subtests with low heterogeneity, so MCV is probably not worth doing.
291:
Both Hart et al studies (2010, 2013) are missing from the reference list. You should check for more missing references.
It seems to me that research based on extended relations would be theoretically superior, not less robust. I could not check the Hart reference to see if some justification was given in that for this claim. Explain?
303:
You should conduct a test to see if this difference is reliable. Fluke perhaps? It goes against all standard theory that h2 should be lower for whites.
338:
Actually, the three way pattern of results is exactly reverse. In sibs: W>H>B. In extended, B>H>W. Very strange. Analytic error? I note that it is found using both methods of dealing with age from looking at Table 12.
353:
Sampling error does not seem a plausible candidate. All sample sizes are in the thousands. From looking at Table 12, the smallest sample is N=1112. Very large for a BG study.
367:
The extended model, does that also use siblings, or only cousin and further away data? Perhaps try estimates based only on e.g. cousins.
I uploaded a new edition for D's preview. I made most of the corrections suggested by Emil. I'm not sure how to search for spaces using my version of Word (using " " doesn't work), so I will have to deal with that another time. I replied in detail to Emil's comments but just before positing my computer shut down. I don't have time now to rewrite the reply. I will let D handle most of the replies, as I will be out of town for the next week.
One thing, concerning the CNLSY analysis, Joe Rodgers (personal communication, December 02, 2013) told us the following:
We followed Joe's advice and reported both sibling and extended results and offered a possible explanation. One explanation not discussed is epigenetics, which has a differential effect on siblings and cousins. Omri Tal provided a model to determine variance due to epigeneticd which, I believe, could be applied to the CNLSY data, but doing so would require the writing of an R program which is beyond both my capability and patience.
Another, Emil said: "You should conduct a test to see if this difference is reliable. Fluke perhaps? It goes against all standard theory that h2 should be lower for whites."
As I explained to Malloy, D, and others this "standard theory" supposes that scores for minority groups are environmentally depressed. If they are not -- if the scores are genetically depressed -- this same theory predicts equal ACE such as found within populations between the upper and lower percentiles e.g., 15th percentile versus 85%.
Later.
One thing, concerning the CNLSY analysis, Joe Rodgers (personal communication, December 02, 2013) told us the following:
"John -- I have a few comments. First, it's pretty much an empirical question as to how the cousins work within the modeling. Sometimes they help a great deal, sometimes they mess things up -- it's a little post hoc, of course, but there are reasons we can develop to explain each. As you note, the potential violation of the EEA is one of those.
So over many years, our teams' standard strategy has been to fit models using cousins, and also not using cousins. Most often, it doesn't matter; sometimes it does. In the first case, we report the result with the larger sample size, and note the other finding as well. In the latter, we report one or both, and interpret the potential problem (usually through violation of the EEA). These statements have applied to both the NLSY79 and NLSYC samples. In cross-generational analysis (Rodgers et al, Behavior Genetics, 2008), the EEA takes on a different status, as we discuss in that paper -- it's attached.
I'm attaching three additional papers that show different outcomes in regards the cousin -- see p. 380 in Rodgers, Rodgers, & Li; p. 37 in Rodgers, Rowe, & Buster, & page 356 in Rodgers, Bard, & Miller.
It's worth noting that Fisher talked about cousins as a "problem category" -- I have that cite somewhere, but I can't easily put my hands on it. He found the cousins often had higher kinship correlations that quantitative genetic models suggested they should -- which is the finding that we usually get if they're problematic.
You can follow our typical approach, or not -- and if you do, you have some papers to cite. But at the bottom line, it's up to each research team to make this call, in relation to the EEA assumption, in relation to sample size considerations, whether the DV being used seems plausible in terms of the EEA assumption, etc., etc.
Good luck with this -- hope the comments above are helpful."
We followed Joe's advice and reported both sibling and extended results and offered a possible explanation. One explanation not discussed is epigenetics, which has a differential effect on siblings and cousins. Omri Tal provided a model to determine variance due to epigeneticd which, I believe, could be applied to the CNLSY data, but doing so would require the writing of an R program which is beyond both my capability and patience.
Another, Emil said: "You should conduct a test to see if this difference is reliable. Fluke perhaps? It goes against all standard theory that h2 should be lower for whites."
As I explained to Malloy, D, and others this "standard theory" supposes that scores for minority groups are environmentally depressed. If they are not -- if the scores are genetically depressed -- this same theory predicts equal ACE such as found within populations between the upper and lower percentiles e.g., 15th percentile versus 85%.
Later.
I mistyped that. I meant higher for whites, lower for whichever underperforming minority the PC people feel bad for.
I made two more minor corrections plus those below.
"Dalliard's name is misspelled."
Fixed.
"Needs a cite for claims. Pick one of the recent review articles. Last sentence should probably get "in adults"."
Citation at end of paragraph.
"You mean "genetic" (broad). E includes error, which is worth mentioning because authors rarely correct for test unreliability so their results are underestimates as pointed out by Jensen in his '69 paper."
Fixed.
"h2 isnt the same as a2. a2 refers to additive genetic factors. h2 may or may not include more depending on whether it is broad or narrow. Narrow h2 is the same as a2."
h^2 typically means narrow heritability, while broad heritability is H^2.
http://books.google.com/books?id=0jB9NzDV4hUC&pg=PA256&lpg=PA256&dq=%22broad+heritability%22+H2&source=bl&ots=s_h7fPx72k&sig=rP3ojCOYRH4TrTXZc_qlFlk3nB0&hl=en&sa=X&ei=__zjU7rgFMKsyQSCyIGoBw&ved=0CDsQ6AEwBQ#v=onepage&q=%22broad%20heritability%22%20H2&f=false
"This overlaps with the beginning of the Introduction. Rewrite."
Fixed.
"Odd phrasing. “Meta-analytic method” maybe? Or just “Method”"
Added colon. “Meta-analysis: method
"Desegregated sounds like you are saying they live in racially segregated communities as in Apartheid times."
Fixed.
"I don't understand the results in Table 7. There are clearly non-zero values in Unstandardized col. E, but they are exactly 0 in Standardized. Explain?"
Added above: "Regarding the computation of ACE estimates, following ordinary practice, we standardized the values such that the total variance added up to 1.00 and no A, C, or E values were negative."
"Where is “the first table”? I can't find a table with results for the g factor."
rewrote as: "g-scores were derived from scores on Digit Span Forward and Backward, PIAT-M, PIAT-RR, PIAT-RC, and PPVT tests taken at age 11. The results are shown in the supplementary file. The genetic variances were 0.61 for Whites, 0.55 for Blacks, and 0.60 for Hispanics. The fact that the g scores are based on multiple tests should make the results reasonably reliable."
"Both Hart et al studies (2010, 2013) are missing from the reference list. You should check for more missing references."
Fixed.
"It seems to me that research based on extended relations would be theoretically superior, not less robust. I could not check the Hart reference to see if some justification was given in that for this claim. Explain?"
Discussed above.
"The extended model, does that also use siblings, or only cousin and further away data? Perhaps try estimates based only on e.g. cousins."
Extended = sibs + cousins. The cousins only shows very high h^2.
"Dalliard's name is misspelled."
Fixed.
"Needs a cite for claims. Pick one of the recent review articles. Last sentence should probably get "in adults"."
Citation at end of paragraph.
"You mean "genetic" (broad). E includes error, which is worth mentioning because authors rarely correct for test unreliability so their results are underestimates as pointed out by Jensen in his '69 paper."
Fixed.
"h2 isnt the same as a2. a2 refers to additive genetic factors. h2 may or may not include more depending on whether it is broad or narrow. Narrow h2 is the same as a2."
h^2 typically means narrow heritability, while broad heritability is H^2.
http://books.google.com/books?id=0jB9NzDV4hUC&pg=PA256&lpg=PA256&dq=%22broad+heritability%22+H2&source=bl&ots=s_h7fPx72k&sig=rP3ojCOYRH4TrTXZc_qlFlk3nB0&hl=en&sa=X&ei=__zjU7rgFMKsyQSCyIGoBw&ved=0CDsQ6AEwBQ#v=onepage&q=%22broad%20heritability%22%20H2&f=false
"This overlaps with the beginning of the Introduction. Rewrite."
Fixed.
"Odd phrasing. “Meta-analytic method” maybe? Or just “Method”"
Added colon. “Meta-analysis: method
"Desegregated sounds like you are saying they live in racially segregated communities as in Apartheid times."
Fixed.
"I don't understand the results in Table 7. There are clearly non-zero values in Unstandardized col. E, but they are exactly 0 in Standardized. Explain?"
Added above: "Regarding the computation of ACE estimates, following ordinary practice, we standardized the values such that the total variance added up to 1.00 and no A, C, or E values were negative."
"Where is “the first table”? I can't find a table with results for the g factor."
rewrote as: "g-scores were derived from scores on Digit Span Forward and Backward, PIAT-M, PIAT-RR, PIAT-RC, and PPVT tests taken at age 11. The results are shown in the supplementary file. The genetic variances were 0.61 for Whites, 0.55 for Blacks, and 0.60 for Hispanics. The fact that the g scores are based on multiple tests should make the results reasonably reliable."
"Both Hart et al studies (2010, 2013) are missing from the reference list. You should check for more missing references."
Fixed.
"It seems to me that research based on extended relations would be theoretically superior, not less robust. I could not check the Hart reference to see if some justification was given in that for this claim. Explain?"
Discussed above.
"The extended model, does that also use siblings, or only cousin and further away data? Perhaps try estimates based only on e.g. cousins."
Extended = sibs + cousins. The cousins only shows very high h^2.
Attached is our latest -- and hopefully final -- version. I corrected some typos in the tables and Dalliard corrected some language issues. I also attached Hart et. al.'s (2014) data.
23: sp
46: sp
67: sp
95: sp
102: sp
122: sp
126: sp
137: sp
148: sp
158:
“Generally, the B/W and H/W mean score differences in these samples were somewhat smaller than those typically found in the national population at large.”
This would indicate that perhaps some of the lower tail blacks are missing from the studies, yes? If those lower tail blacks also have the worst environments, then this can bias the h^2 estimate upwards for blacks. I know you only have a few samples, but theoretically, you can test this using r(W-B h^2 x W-B IQ d). If the depression hypothesis is right, this should be negative.
Alternatively, perhaps the W-B d is just smaller now than it used to as claimed by some. Or not fully formed at these ages.
160: sp
216: sp
217: sp
264:
“Another assumption is that no assortative mating took place in the parental generation with respect to cognitive abilities.”
Is it possible to run the model with estimates of assortative mating?
Vandenberg, S. G. (1972). Assortative mating, or who marries whom?. Behavior genetics, 2(2-3), 127-157.
Cites some studies. Range is .44 to .60. Unweighted mean is 48.75. It seems foolish to conduct a study assuming a zero value for something that is known to be substantial. On the other hand, I'm not aware of any studies of assortative mating by race. If the strength of this is different, it should change the amount of genetic variance within race. This influences d's values.
272: sp
328: sp
339: sp
340: sp
359: sp
361: sp
421: sp
426: sp
433: sp
478: sp
46: sp
67: sp
95: sp
102: sp
122: sp
126: sp
137: sp
148: sp
158:
“Generally, the B/W and H/W mean score differences in these samples were somewhat smaller than those typically found in the national population at large.”
This would indicate that perhaps some of the lower tail blacks are missing from the studies, yes? If those lower tail blacks also have the worst environments, then this can bias the h^2 estimate upwards for blacks. I know you only have a few samples, but theoretically, you can test this using r(W-B h^2 x W-B IQ d). If the depression hypothesis is right, this should be negative.
Alternatively, perhaps the W-B d is just smaller now than it used to as claimed by some. Or not fully formed at these ages.
160: sp
216: sp
217: sp
264:
“Another assumption is that no assortative mating took place in the parental generation with respect to cognitive abilities.”
Is it possible to run the model with estimates of assortative mating?
Vandenberg, S. G. (1972). Assortative mating, or who marries whom?. Behavior genetics, 2(2-3), 127-157.
Cites some studies. Range is .44 to .60. Unweighted mean is 48.75. It seems foolish to conduct a study assuming a zero value for something that is known to be substantial. On the other hand, I'm not aware of any studies of assortative mating by race. If the strength of this is different, it should change the amount of genetic variance within race. This influences d's values.
272: sp
328: sp
339: sp
340: sp
359: sp
361: sp
421: sp
426: sp
433: sp
478: sp
23: sp 46: sp 67: sp 95: sp 102: sp 122: sp 126: sp 137: sp 148: sp 160: sp 216: sp 217: sp 272: sp 328: sp 339: sp 340: sp 359: sp 361: sp 421: sp 426: sp 433: sp 478: sp
When I search for spacing with word, I don't get the same number of hits. Also, your line numbers don't match with those I get, so are of no help. Anyways, I found a few more and corrected those.
Alternatively, perhaps the W-B d is just smaller now than it used to as claimed by some. Or not fully formed at these ages.
With regards to r (B/W d x B/W h^2), age and test type were major confounds; there weren't enough samples to control for these. The d-values averaged to 0.88 for the 6 non CNLSY samples. Age and test type were likely factors; birth year wasn't. Regarding Rowe and Cleveland's CNLSY sample, the d values were only 0.3 to 0.5, but this had more to do with the waves selected and the nature of the achievement tests used. Murray found no birth cohort effect for this sample.
To clarify, I added:
"Generally, the average H/W mean score difference (d-value) in these samples was of a similar size to that found nationally (0.7 versus 0.7), while the average B/W d-value was somewhat smaller (0.8 versus 1.0). The reduced magnitude of the B/W d-value from these samples was likely due to a participant age and test type effect....
Were a bio-ecological model correct, one might expect that d-values would positively correlate with heritability differences, such that when d-values were larger, the lower scoring population would show more depressed test heritabilities. Unfortunately, our samples do not allow us to robustly determine whether this is the case as they differ in participant age and test type, differences which would cofound any such analysis and which cannot be controlled for given the dearth of samples available."
I'm not aware of any studies of assortative mating by race. If the strength of this is different, it should change the amount of genetic variance within race.
I'm not either.
I've just had a really quick look at your paper. I will have a more deep reading in the coming days. In the first paragraph you wrote "In behavioral genetics, the sources of IQ variance can be partitioned into three components: additive heredity (h2), shared environment (c2), and unshared environment (e2)".
However, models sometimes also include D, which indicates dominance (non-additive) effects. It's well known that part (albeit not the majority) of the variance in IQ is also due to non-additive effects.
However, models sometimes also include D, which indicates dominance (non-additive) effects. It's well known that part (albeit not the majority) of the variance in IQ is also due to non-additive effects.
I've just had a really quick look at your paper. I will have a more deep reading in the coming days. In the first paragraph you wrote "In behavioral genetics, the sources of IQ variance can be partitioned into three components: additive heredity (h2), shared environment (c2), and unshared environment (e2)".However, models sometimes also include D, which indicates dominance (non-additive) effects. It's well known that part (albeit not the majority) of the variance in IQ is also due to non-additive effects.
I'm not sure by D changed "genetic influence" back to "additive heredity". As it is, we used Falconer's formula in which heritability is broad sense e.g., h2(b) = 2(rmz - rdz). So this should read: "genetic influence (broad heritability) (h2), shared environment (c2), and unshared environment (e2)". But maybe he had something else in mind.
Falconer's forumula gives an estimate of broad sense heritability, which is A+D. So it's better to use h2 than A, because A is only additive. Just specify that h2= A+D
I opened the file with LibreOffice. You could download that and do the same. It's free. I tried opening it with Word 2010 too. The line numbers are off, but only by 1, and not initially. Almost all the sp concern extra spaces at the end of paragraphs for no reason.
It's untrue that ACE model specifies additive heritability. It's not even clear if some experts understand it well, or not. Here's an illustration :
Haworth, C. M. A., Wright, M. J., Luciano, M., Martin, N. G., de Geus, E. J. C., van Beijsterveldt, C. E. M., ... & Plomin, R. (2010). The heritability of general cognitive ability increases linearly from childhood to young adulthood. Molecular Psychiatry, 15, 1112-1120.
Now, click on it, do CTRL+SHIFT+F, type "additive". You get 4 hits. Now, read these sentences.
How is it wrong ? Well, I have written an article on income mobility where, at some point, I talked about "genoeconomics". Here's the paragraph that will help to understand this problem :
In other words, before talking about additive heritability, you must test ACE against ADE. If the D component is clearly larger than 0, you probably have to select it, and then look at the A component. But unless you test models vs models, you can't have any certainty about what is additive and what is not.
The A in ACE is just an assumption, and it must be tested to see if that holds or not.
Haworth, C. M. A., Wright, M. J., Luciano, M., Martin, N. G., de Geus, E. J. C., van Beijsterveldt, C. E. M., ... & Plomin, R. (2010). The heritability of general cognitive ability increases linearly from childhood to young adulthood. Molecular Psychiatry, 15, 1112-1120.
Now, click on it, do CTRL+SHIFT+F, type "additive". You get 4 hits. Now, read these sentences.
These genetic and environmental effects are commonly represented as A, C and E. ‘A’ is the additive genetic effect size, also known as narrow heritability.
How is it wrong ? Well, I have written an article on income mobility where, at some point, I talked about "genoeconomics". Here's the paragraph that will help to understand this problem :
Similarly, Hyytinen et al. (2013) showed, using the Older Finnish Twin Cohort Study for Finland, with MZ and DZ pairs of 620 and 1146 for women and 494 and 1094 for men, that 24% and 54% of variance in lifetime income for women and men, respectively, is due to genetic factors, whereas shared environment is small. The authors begin to cite studies showing that schooling reforms have enhanced earnings mobility. As explained above, this is irrelevant due to its weak intergenerational transmission through the family.
The income data was from administrative registers and thus do not suffer measurement errors due to misreporting. For their analysis, they use the Defries-Fulker regression method, which can be formulated as INC1=β0+β1INC2+β2R+β3(INC2*R)+ϵ, where INC is income, β0 is the intercept, β1 is a measure of shared environment (C), β2 is a coefficient of genetic relatedness (r=1.00 for MZ and r=0.50 for DZ twins, and thus assumes full additivity), β3 the heritability, ϵ the error term (E) which includes both the non-shared environment as well as measurement error. It is possible to include a parameter for non-additive genetic effect, or dominance (D), which in this case can be denoted as β4(INC2*D), where D=1.00 for MZ and D=0.25 for DZ twins. This can be called the ADE model, where β3 and β4 evaluate the additive (i.e., narrow) and non-additive heritability for income. The sum of the additive (A) and non-additive (D) is called the broad-sense heritability. Thus the difference between ACE and ADE models is that the former, but not the latter, assumes the heritability is purely additive.
We read the results from their model fitting in table 3. For females, shared environment (C) is small (0.10) in the ACE model whereas for males the C parameter is negative and this is indicative of dominance (D) effect, since C is the mirror of D, and conversely. For both gender group, based on the model fit index Akaike (AIC) the AE has the worst fit but the ACE and ADE have equal fit. At first glance, one believes it is impossible to select among them. However, given that D is negative for women, this model is probably ill-specified and thus they have (rightly) opted for the ACE. Because C is negative in ACE and this in turn is suggestive of large D parameter, which is confirmed in ADE model, the authors have correctly chosen the ADE as the preferred model for men. In the ACE, women and men have a heritability of 0.24 and 0.77. In ADE, the A and D parameters for women amount to 0.54 and -0.20 and for men to 0.07 and 0.47. The sum of A and Z gives a broad heritability of 0.54 for men. Given that ACE should be preferred for females, their heritability is 0.24. Interestingly, they note (table A2) that the use of a broader measure of income (which includes capital income and transfers, such as unemployment benefits and parental leave benefits) has improved the heritability estimates, which were 0.42 and A+D=0.26+0.33=0.59 for women and men, given their respective preferred model, AE and ADE. This provides another illustration of how measurement errors can reduce heritability.
When they attempt to add education as covariate in the regression, thus holding education constant, the parameter estimates (A,C,D,E) remain somewhat unchanged compared to table 3, even though heritability has diminished somewhat. When education effect is deducted from income, the A, C, D and E parameters appear similar.
In other words, before talking about additive heritability, you must test ACE against ADE. If the D component is clearly larger than 0, you probably have to select it, and then look at the A component. But unless you test models vs models, you can't have any certainty about what is additive and what is not.
The A in ACE is just an assumption, and it must be tested to see if that holds or not.
The A in ACE is just an assumption, and it must be tested to see if that holds or not.
But practically speaking, can I use "A" in charts that include h^2 estimates derived from Falconer's formula? Or do I have to change this to h^2 (and defined h^2 as "genetic influence")? Specifically, are the charts I have fine in regards to the ACE labels.
My recommendation is that you should precise what study is from falconer's calculation, and what is from ACE modeling, after you have precised that the studies of race-ACE interaction did not test ACE vs ACDE (or ADE) models.
For example, in your tables, use only h2 in the columns, and just below the tables, add a note and put * for studies using falconer and ** for studies of ACE modeling, and precise that in the case where you have ** the h2 referred to the A component of ACE.
For example, in your tables, use only h2 in the columns, and just below the tables, add a note and put * for studies using falconer and ** for studies of ACE modeling, and precise that in the case where you have ** the h2 referred to the A component of ACE.
Falconer's formula does not give H2. Falconer and Mackay (1996) say that it gives some value that is closer to H2 than h2 but it's still not H2.
ACE or ADE model fitting to twin data is limited in that C and D components cannot be simultaneously estimated. Moreover, assortative mating (AM) is not considered. What this means is that the estimates from these models cannot be correct. ACE or AE models usually show the best fit to IQ data from twins, but that's in part because the biases due to non-additivity and AM have opposite effects. Non-additivity increases the MZ correlation more than the DZ correlation, but AM increases the DZ correlation without increasing the MZ correlation, so these effects tend to cancel each other out. In practice, the A estimate from ACE or AE models may be close to H2. <a href="http://link.springer.com/article/10.1007/s10519-011-9507-9/fulltext.html">Vinkhuyzen et al.</a> had data where both D and AM could be estimated, and they conclude that much of IQ heritability is non-additive but their H2 estimate comes to about 80 percent, which is similar to h2 estimates from studies that assume only additive heredity and no AM.
Because almost all IQ studies use ACE or AE models and Falconer's formula does not give H2, either, I think we should talk about h2 or A rather than H2 in the paper. We can mention the methods used in each study.
Also, MH, 'precise' is not a verb in English. Use 'clarify' or 'specify' instead.
ACE or ADE model fitting to twin data is limited in that C and D components cannot be simultaneously estimated. Moreover, assortative mating (AM) is not considered. What this means is that the estimates from these models cannot be correct. ACE or AE models usually show the best fit to IQ data from twins, but that's in part because the biases due to non-additivity and AM have opposite effects. Non-additivity increases the MZ correlation more than the DZ correlation, but AM increases the DZ correlation without increasing the MZ correlation, so these effects tend to cancel each other out. In practice, the A estimate from ACE or AE models may be close to H2. <a href="http://link.springer.com/article/10.1007/s10519-011-9507-9/fulltext.html">Vinkhuyzen et al.</a> had data where both D and AM could be estimated, and they conclude that much of IQ heritability is non-additive but their H2 estimate comes to about 80 percent, which is similar to h2 estimates from studies that assume only additive heredity and no AM.
Because almost all IQ studies use ACE or AE models and Falconer's formula does not give H2, either, I think we should talk about h2 or A rather than H2 in the paper. We can mention the methods used in each study.
Also, MH, 'precise' is not a verb in English. Use 'clarify' or 'specify' instead.
© Copyright 2024 OpenPsych - All Rights Reserved