Hello There, Guest!  
 Previous 1 2 3 4 5 9 Next   

[OBG] Sexual selection explains sex and country differences in fluid g

#21
(2014-Jul-06, 19:55:55)Barleymow Wrote: This is an interesting paper which I hope to eventually see published. However, a couple of suggestions. (1) I wonder if height is a good measure to use. These are often self-estimates and are thus not very reliable. As such, it might be better to use a number of different estimates to obviate the unreliability. For example, Lynn did a paper on penis length in different European countries. (2) It strikes me that the paper would be strengthened by using all of the PISA data as there is a high correlation between this and IQ.


The penis data was made up. Lynn apparently found them on the internet and didn't check the source. Philbrick pointed this out to me in an earlier review of a paper I wrote, I think.
 Reply
#22
When someone performs multiple regression, especially in case where there is few indep var, I always recommend to test for the presence of interaction (i.e., non linear relationship.

When you try the following :

COMPUTE sex_gdp_interaction=Difference*GDP.
EXECUTE.

REGRESSION
/DESCRIPTIVES MEAN STDDEV CORR SIG N
/MISSING LISTWISE
/STATISTICS COEFF OUTS CI(95) R ANOVA COLLIN TOL CHANGE ZPP
/CRITERIA=PIN(.05) POUT(.10)
/NOORIGIN
/DEPENDENT PISA_Score_Total
/METHOD=ENTER Difference GDP sex_gdp_interaction
/PARTIALPLOT ALL
/SCATTERPLOT=(*ZRESID ,*ZPRED)
/RESIDUALS DURBIN HISTOGRAM(ZRESID) NORMPROB(ZRESID).

The regress coeff for sex, gdp, and interaction are 0.170, 0.426, 0.185. It's clearly different from what you initially have (0.323, and 520). The gender effect has been divided by an half.

When you repeat the same procedure with pisa SD in dependent var., instead of having -0.490 and -0.147 for sex and gdp, you have -0.302, -0.030, -0.226 for sex, gdp, and their interaction, respectively. This means there is virtually no main effect of GDP, and its (direct) effect is only a matter of interaction, (i.e., the effect of GDP, while probably null at low value of sex difference, it increases as the sex difference increases).

Practioners should also and always check for normal distribution of the residuals. Hopefully, they look normal.

MR, in any case, is a sub-optimal way to "control" for covariates (or for what it means). If you need to look at the effect of GDP net of that of gender, you should better use male/female PISA as dependent var, separately, but if that means you have only 1 indep var, MR and bivariate correlation are not distinguishable methods (they produce the same results). For what I have seen, using SD or score for male/female separately makes no difference for the correlation with GDP.
 Reply
#23
(2014-Jul-09, 17:40:15)menghu1001 Wrote: When someone performs multiple regression, especially in case where there is few indep var, I always recommend to test for the presence of interaction (i.e., non linear relationship.

When you try the following :

COMPUTE sex_gdp_interaction=Difference*GDP.
EXECUTE.

REGRESSION
/DESCRIPTIVES MEAN STDDEV CORR SIG N
/MISSING LISTWISE
/STATISTICS COEFF OUTS CI(95) R ANOVA COLLIN TOL CHANGE ZPP
/CRITERIA=PIN(.05) POUT(.10)
/NOORIGIN
/DEPENDENT PISA_Score_Total
/METHOD=ENTER Difference GDP sex_gdp_interaction
/PARTIALPLOT ALL
/SCATTERPLOT=(*ZRESID ,*ZPRED)
/RESIDUALS DURBIN HISTOGRAM(ZRESID) NORMPROB(ZRESID).

The regress coeff for sex, gdp, and interaction are 0.170, 0.426, 0.185. It's clearly different from what you initially have (0.323, and 520). The gender effect has been divided by an half.

When you repeat the same procedure with pisa SD in dependent var., instead of having -0.490 and -0.147 for sex and gdp, you have -0.302, -0.030, -0.226 for sex, gdp, and their interaction, respectively. This means there is virtually no main effect of GDP, and its (direct) effect is only a matter of interaction, (i.e., the effect of GDP, while probably null at low value of sex difference, it increases as the sex difference increases).

Practioners should also and always check for normal distribution of the residuals. Hopefully, they look normal.

MR, in any case, is a sub-optimal way to "control" for covariates (or for what it means). If you need to look at the effect of GDP net of that of gender, you should better use male/female PISA as dependent var, separately, but if that means you have only 1 indep var, MR and bivariate correlation are not distinguishable methods (they produce the same results). For what I have seen, using SD or score for male/female separately makes no difference for the correlation with GDP.


The linear model is the most parsimonious and probably to be preferred. It's likely that any interaction effects are due to statistical flukes.
Also throwing in an interaction effect between sex and gdp seems a bit sketchy, without a theoretical reason for doing so.However, I appreciate your help and am not denying its validity.I just think it's not entirely appropriate for this analysis.
 Reply
#24
(2014-Jul-09, 16:51:13)Emil Wrote:
(2014-Jul-06, 19:55:55)Barleymow Wrote: This is an interesting paper which I hope to eventually see published. However, a couple of suggestions. (1) I wonder if height is a good measure to use. These are often self-estimates and are thus not very reliable. As such, it might be better to use a number of different estimates to obviate the unreliability. For example, Lynn did a paper on penis length in different European countries. (2) It strikes me that the paper would be strengthened by using all of the PISA data as there is a high correlation between this and IQ.


The penis data was made up. Lynn apparently found them on the internet and didn't check the source. Philbrick pointed this out to me in an earlier review of a paper I wrote, I think.


That's correct.
 Reply
#25
(2014-Jul-09, 22:00:15)Duxide Wrote: The linear model is the most parsimonious and probably to be preferred. It's likely that any interaction effects are due to statistical flukes.
Also throwing in an interaction effect between sex and gdp seems a bit sketchy, without a theoretical reason for doing so.However, I appreciate your help and am not denying its validity.I just think it's not entirely appropriate for this analysis.


In general, the most parsimonious model should be preferred when taking into account model fit. In principle, however, when interaction is present, the model fit must necessarily favor a model with interaction. Theoretically, I still think it's quite interpretable. The results above mean that GDP has no main effect and the strength of its effect depend on the size of the gender difference in PISA. Now the whole thing is to interpret this outcome in light of the theory advanced in your article. I'll try to read the text more carefully and see if I get some ideas.
 Reply
#26
(2014-Jul-10, 16:22:54)menghu1001 Wrote:
(2014-Jul-09, 22:00:15)Duxide Wrote: The linear model is the most parsimonious and probably to be preferred. It's likely that any interaction effects are due to statistical flukes.
Also throwing in an interaction effect between sex and gdp seems a bit sketchy, without a theoretical reason for doing so.However, I appreciate your help and am not denying its validity.I just think it's not entirely appropriate for this analysis.


In general, the most parsimonious model should be preferred when taking into account model fit. In principle, however, when interaction is present, the model fit must necessarily favor a model with interaction. Theoretically, I still think it's quite interpretable. The results above mean that GDP has no main effect and the strength of its effect depend on the size of the gender difference in PISA. Now the whole thing is to interpret this outcome in light of the theory advanced in your article. I'll try to read the text more carefully and see if I get some ideas.


I am afraid you'll have to reread my paper because I've substantially enlarged and updated the introduction with theory, slighly changed discussion and added to methods section. I attach the updated version.


Attached Files
.docx   OPEN BG Sexual selection.docx (Size: 40.93 KB / Downloads: 539)
 Reply
#27
The paper is improving. Here are my comments on the latest draft.

Line 17:
"Sexual dimorpshism"

Spelling.

30:
"However, sexual selection raises the average phenotypic trait value not only in the selected sex, but to a lesser extent also in the opposite sex, via the mechanism of genetic correlation between homologous characters of the sexes (that is, the correlation between the additive effects of genes as expressed in males and females) (Lande, 1980)."

Rewrite.

34:
"strenght"

Spelling.

42:
"a X-linked"

Grammar.

51:
"reciprocral crosses"

Spelling.

77:
"dimorphism.The"

Spacing.

"strenght"

Sp.

89:
"Van Homrigh et al.,2007"

Spacing.

90:
"fixed.Thus"

Spacing.

96:
" Piffer (2014)"

Typo.

You should perhaps note the correlation between CPS and IQ at the country level. If there are any personal level studies, cite them. I don't know any.

160:
"(r= - .484; p= 0.01; N= 41)"

Why are 3 cases missing?

164:
Headers in tables should be bolded. "OECD" is not a header, but a category of countries. The usual practice is italicization of categories in lists.

168:
Spacing in table isn't very optimal for headers.

199:
"The correlation reached significance after removing the effect of GDP, suggesting that economic development masks the relationship between sex dimorphism in IQ and country IQ scores through its negative effect on male intellectual advantage and its positive association with country IQ."

It is better to say that the effect size increased after partialling out GDP. Whether a p-value is below some arbitrary number is not important.

209:
"Supporting the brawn vs brain evolutionary scenario, male height was found to be negatively related to sex differences in intelligence, which is a proxy for sexual selection strength. This suggests that there is a trade-off in sexual selection between physical power or attractiveness and intellectual abilities. This provides a possible explanation for the finding by Piffer (2014) that frequencies of alleles known to increase height had a strong inverse correlation between populations to frequencies of alleles that increase IQ"

It seems worth nothing that this is apparently an inter-population effect, as within populations g and height are known to correlate about .2. How does the author propose that these two contradictory tendencies be interpreted?

215:
"The results of this study have multiple implications. First, the evolution of intelligence has been probably affected by evolutionary forces that acted differently on males and females. Although with the present data it is impossible to determine the precise mechanism (i.e. whether it is was due to female choice or higher reproductive success of high IQ males via higher wealth and social status) this study provides encouraging results for future investigations into the role played by sexual selection on intelligence during prehistoric and historic times. Another implication of this study is that intelligence has continued to evolve after different human populations migrated out of Africa and possibly up to the 19th century, as suggested by the substantial variability in sex differences even between neighbouring countries."

This section should be worded more cautiously. The results have yet to be replicated and the PISA sample is not so diverse, mostly countries with Europeans in them.

224:
"Finally, the failure of GWAS to find genes accounting for a significant variation in intelligence could be due to their exclusive focus on the autosomal genome and the findings presented in this paper could provide a rationale for an extension of genomic studies of cognition to the sex chromosomes"

Missing a dot at the end.

The author should cite a paper or two about how GWAS hits for g has generally not replicated. It is also good to mention that recent and larger GWASs have replicated hits of g alleles.
 Reply
#28
Jensen's Educational Differences p. 215 sums up quite a few studies for correlations between same-sex and opposite-sex siblings. There is no difference. To me this indicates that there is no or very little sex-linked heritability.

How would the author deal with this? It means that the basis of the prediction is groundless, even if the prediction turned out to be true. Coincidence?


Attached Files Thumbnail(s)
   
 Reply
#29
A journalist asked Einstein what he would do if Eddington's observations failed to match his theory. Einstein replied : "I would feel sorry for the God lord. The theory is correct".
 Reply
#30
Quote:Dear Davide Piffer,

Thank you for your submission to Evolutionary Psychology. We have given your submission full attention. However, after consultation with the Editorial board, we have decided that your manuscript is not suitable for publication in Evolutionary Psychology, and thus won't be sent out for in-depth review. I am sorry for being the bearer of what must be negative news. The Editors of Evolutionary Psychology aim to give quick feedback particularly with submissions, which are unlikely to get accepted even after in depth review and/or revision. Alas your submission falls into this category and was therefore rejected at this stage.

Best of luck with your work.

Sincerely,

Bernhard Fink
 Reply
 Previous 1 2 3 4 5 9 Next   
 
 
Forum Jump:

Users browsing this thread: 1 Guest(s)