This paper is a replication of "Evidence for a Paternal Age Effect on Leftism" without dichotomization and the result here looks more convincing than last time. The variables of interest now contain more questionnaires related to each dimension (LGBT, BLM, feminism), which greatly improves upon the last study. I'm going to focus mainly on the presentation and method.

The first surveys 1175 white American men and gives them the general leftism test, and asks their father’s age when they were born.

I highly recommend displaying mean and SD of age for this sample. Same goes for the second survey with fathers over 50 yrs old.

These dimensions are hypothesized to be common to empire decline, and covary due to being the result of mutational pressure on the same genes.

Need references.

By pre-hoc design, the three factors were summed and a general factor was derived by varimax factor analysis on these sums.

The sentence is very confusing. These three factors I believe refer to LGBT, BLM, feminism. Factor is a term used for latent variable, not observed variable. If you summed R1-R6, G1-G6, F1-F6 before factor analyzing them, you should refer to them not as factors but as sum scores. Also, you should check whether the distribution of these three sum scores is normal prior to factor analysis. 

Now, about "varimax factor analysis". First, I understand what you mean, but it needs to be properly written, e.g., factor analysis with varimax rotation. Second, I wish the choice of the rotation method is discussed with respect to varimax because this rotation specifies that the factors (i.e., LGBT, BLM, feminism factors) are uncorrelated. Refer to the discussion in the following paper:
Fabrigar, L. R., Wegener, D. T., MacCallum, R. C., & Strahan, E. J. (1999). Evaluating the use of exploratory factor analysis in psychological research. Psychological Methods, 4, 272–299.

Also, I wouldn't use rotation if the purpose is to get a general factor score, especially from a one-factor solution, to be used in a regression analysis. I'm not sure I understand the purpose of it. 

The full data will be publicly available on the author’s Github if you wish to verify this.

Never use "you" in a paper. 

Figures 1-4 and 7-8 are not cited/discussed in the main text, the figures are just merely presented out of nowhere. 

The title in figures 5-6 ends with a dot, but only for figures 5-6. Better be consistent: using it for all figures or none. Also, Figure 6 is actually a table. 

Study 1 subsection must appear under Result section. Regarding the result section, I wish there is a general presentation of the result for both study 1 and study 2. As it stands now, it feels like a low effort presentation of the result section. For instance, Figures 4 and 5 are described with two short sentences. No mention of numbers of interest (especially Figures 5 and 7). I know the readers can zoom in to check the numbers in the figures, but the author is accountable for presenting the result properly, in a more academic way.

About this table, you included an interaction variable (age*paternal age) without mentioning it in the main text. Please provide an explanation as to why you want to use this interaction variable. In your earlier study, you did not include this interaction variable. Moreover, the notes under the table indicates you use robust standard errors. This is another detail not mentioned in the text and it should be made clear why you opted for robust SE. I also don't understand this sentence here: 

We find that in the multiple regression model, the correlation of leftism with paternal age decreased by less than 0.01

The number is not what I see in the table. And shouldn't the relationship be positive and not negative? Also, it is more accurate to refer to the effect as beta regression β instead of correlation, to avoid confusion. Eventually, mention in the main text the p value or, better, the confidence intervals along with the β estimate.

For IQ, which likely has a similar, but orthogonal, genetic structure to conservatism, a paternal age correlation of a similar magnitude has been found [9]. This did not weaken when controlled for maternal age, but the analysis lacked the power to properly control for birth order

I would rewrite the sentence as, e.g., "... of a similar magnitude has been found by Wang [9]. In their study, this did not weaken..." since the entire paragraph speaks about Wang's study rather than the current study. I find it clearer.

Based on the results, we conclude that there is compelling evidence for a paternal age effect for leftism. The next step is molecular confirmation. Studies which confirm the role of de novo mutation...

The conclusion (or discussion) displays only one short sentence on the results. This is a meager presentation.

Mental illness correlates with leftism on other scales

Provide a reference.

The decline of asabiyyah [5] seems to be a general feature of empire decline.

For convenience, define asabiyyah in the paper, even briefly.

F2. The country would be better if women couldn't vote. (-1)
Age allows one to estimate the base mutational load of an individual's generation while
8. Angrist, J. D., & Pischke, J. S. (2009). Mostly harmless econometrics: An empiricist's

Make sure the marks are curly, not straight, for consistency. Other single marks are all curly.

Bronski, J (2023) Quantitative Sociobiology Manuscript Beta 1.4. Pg. 48

There should be a dot after J, and after 48. Speaking of this reference, it doesn't appear in the main text. I also wasn't able to find this reference. The closest I found was this article (Beta 1.3 however, not 1.4): https://www.josephbronski.com/p/quantitative-sociobiology-manuscript. Unfortunately, it is restricted to paid subscriber, which means a normal reader, even a scholar, won't be able to access it. Perhaps if you want to cite the reference, you can cite the passage at page 48 that you mention, so that readers know exactly the point you have made here without the requirement of accessing the whole article.

The next step for a future study replication should use a more direct question for the variable of left-right ideology such as political views/party.

All in all I have no problem with the methodology of the main analysis, as I once did in your previous article, but the presentation (i.e., write up) needs serious (re)work to make it look more professional. 

1. I highly recommend displaying mean and SD of age for this sample. Same goes for the second survey with fathers over 50 yrs old.

Done

2. Need references.

Added a reference

3. The sentence is very confusing. These three factors I believe refer to LGBT, BLM, feminism. Factor is a term used for latent variable, not observed variable. If you summed R1-R6, G1-G6, F1-F6 before factor analyzing them, you should refer to them not as factors but as sum scores. Also, you should check whether the distribution of these three sum scores is normal prior to factor analysis. 

I changed the wording. They are sum scores. They yield a single general factor. 

I am still reading about factor analysis, does the data have to be normal? For PCA it does not. I added q-q plot r^2s to the paper for the sumscores, basically they are pseudo-normal, they have issues measuring beyond 2 SDs in either direction which will add some noise to all analyses. 

4. Now, about "varimax factor analysis". First, I understand what you mean, but it needs to be properly written, e.g., factor analysis with varimax rotation. Second, I wish the choice of the rotation method is discussed with respect to varimax because this rotation specifies that the factors (i.e., LGBT, BLM, feminism factors) are uncorrelated. Refer to the discussion in the following paper:

I fixed the wording. I’m only using it to find one factor, so varimax producing uncorrelated factors should not impact anything.

5. Also, I wouldn't use rotation if the purpose is to get a general factor score, especially from a one-factor solution, to be used in a regression analysis. I'm not sure I understand the purpose of it. 

It was basically exchangeable with PCA but the varimax factor had a slightly better distribution. But it correlated with the PCA factor at r=.96.

6. Never use "you" in a paper. 

fixed

7. Figures 1-4 and 7-8 are not cited/discussed in the main text, the figures are just merely presented out of nowhere. 

I cited them all now

8. The title in figures 5-6 ends with a dot, but only for figures 5-6. Better be consistent: using it for all figures or none. Also, Figure 6 is actually a table. 

Fixed. 

9. Study 1 subsection must appear under Result section. Regarding the result section, I wish there is a general presentation of the result for both study 1 and study 2. As it stands now, it feels like a low effort presentation of the result section. For instance, Figures 4 and 5 are described with two short sentences. No mention of numbers of interest (especially Figures 5 and 7). I know the readers can zoom in to check the numbers in the figures, but the author is accountable for presenting the result properly, in a more academic way.

Added all of these

10. About this table, you included an interaction variable (age*paternal age) without mentioning it in the main text. Please provide an explanation as to why you want to use this interaction variable. In your earlier study, you did not include this interaction variable. Moreover, the notes under the table indicates you use robust standard errors. This is another detail not mentioned in the text and it should be made clear why you opted for robust SE. I also don't understand this sentence here: 

Robust SE basically does not change the SEs, my OLS code had them on by default. I added stuff on the interaction, it was because Woodley allegedly found an interaction like that with religiosity.

11. We find that in the multiple regression model, the correlation of leftism with paternal age decreased by less than 0.01 The number is not what I see in the table. And shouldn't the relationship be positive and not negative? Also, it is more accurate to refer to the effect as beta regression β instead of correlation, to avoid confusion. Eventually, mention in the main text the p value or, better, the confidence intervals along with the β estimate.

The relationship should be positive with paternal age, positive is the leftist direction, negative with age as older people are less leftist. I changed the wording to make it more clear.

12. I would rewrite the sentence as, e.g., "... of a similar magnitude has been found by Wang [9]. In their study, this did not weaken..." since the entire paragraph speaks about Wang's study rather than the current study. I find it clearer.

Done

13. The conclusion (or discussion) displays only one short sentence on the results. This is a meager presentation.

Added some sentences

14.  Provide a reference. Mental illness correlates with leftism on other scales

Done

15. For convenience, define asabiyyah in the paper, even briefly.

Done

16. Make sure the marks are curly, not straight, for consistency. Other single marks are all curly.

Done

17. There should be a dot after J, and after 48. Speaking of this reference, it doesn't appear in the main text. I also wasn't able to find this reference. The closest I found was this article (Beta 1.3 however, not 1.4): https://www.josephbronski.com/p/quantitative-sociobiology-manuscript. Unfortunately, it is restricted to paid subscriber, which means a normal reader, even a scholar, won't be able to access it. Perhaps if you want to cite the reference, you can cite the passage at page 48 that you mention, so that readers know exactly the point you have made here without the requirement of accessing the whole article.

I removed it since it was uncited

18. The next step for a future study replication should use a more direct question for the variable of left-right ideology such as political views/party.

Issue with this is that such a question is inherently more noisy than a latent factor, like g vs. one specific item, so all correlations will fall, including the heritability and the correlation with paternal age. 

Author has updated the submission to version #2

I read the new version just now, and I'm satisfied with the discussion section (not sure why it's written Conclusion and not Discussion, but it's a nitpick at this point) and generally with the presentation of the result section. We are almost there.

I have a few comments still.

First of all, I have to admit I missed an important detail earlier: the first figure PCA Scree Plot did not have a title (about this one, if you could increase its resolution, this would be excellent). This should have been your Figure 1 (which is also not mentioned or discussed in the main text) so your Figure 2 and following ones should all be re-numbered as well as in the main text.

and asks their father’s age when they were born (mean 61.3, SD 7 years)

It should be: mean = 61.3, SD = 7 years

These dimensions are hypothesized to be common to empire decline, and covary due to being the result of mutational pressure on the same genes [1].

I wished it was a new reference but I guess it's fine too.

a general factor was derived by both factor analysis with varimax rotation.

"both" factor analysis?

This d score is equivalent to an r of about 0.75

In your abstract, d is italicized but not here.

When presenting Figures/Tables in the main text, the typical presentation is that they are called out (refered to) in the main text before, not after the Figure/Table placement. I'm not too picky on this, I'm fine with both, but in your situation, Figure 3 is called out before, while Figure 1-2 are called out after. At least be consistent for all tables/figures. 

Figure 4 shows the correlation between leftism and paternal age. There is a significant positive correlation.

Actually, figure 4 shows only r, CI, and n, but not the p value. Since you wrote that the correlation is "statistically significant" it is odd not to disclose the p value as well.

Figure 5 Correlation matrix for everything.

There should be a dot after Figure 5 (same issue for Table 1). I ask that you reread the paper very carefully, because there were tons of mistakes like these ones. Perhaps I even missed a few more. Also, the title of Figure 5 is not italicized but all other figures/tables are. I don't think the OP journal requires you to italicize the Figure/Table title, and usually don't see these in italics in other journals. So this is very odd. Finally, another nitpick with regard to Figure 5: I think it's better you change the title as follows, e.g.,: Correlation matrix of study variables" because everything is odd here. 

Figure 5 shows all of the correlation coefficients between everything in the data.

Same issue as above; between study variables looks much better.

One star means p < 0.05, two means p < 0.01, and three means p < 0.001.

This is a very unconventional way of reporting the p values. Again I don't want to be very picky and I accept this but I want to warn you that in other journals this may not be accepted. 

the partial correlation of paternal age (.1104)

I mentioned it before but I still believe it's far more accurate, less confusing, to report the estimate as a standardized regression coefficient rather than partial correlation. It's your call though. Don't see it as mandatory.

However, as with the binary measurement from Bronski (2023), we found no evidence of such a correlation.

Since you're not using APA citing style, this should be Bronski [1].

has been found by Wang[8]. 

Spacing.

but it would be nice to verify this directly for leftism as well as IQ

This is another nitpick. After given some thoughts I believe it's best to use formal words for professional papers. So "nice" is a bit odd here. 

Now on your responses:

For PCA it does not. 

In the main text, you said that you attempted both but finally opted for factor analysis with varimax rotation, not PCA, which is why I asked.

I’m only using it to find one factor, so varimax producing uncorrelated factors should not impact anything.

That was still a weird wording for me, because with one factor solution, why rotating?

Robust SE basically does not change the SEs, my OLS code had them on by default. 

Fine, but at least make it clear in the text. Typically, researchers don't use robust SE unless they suspect or know the residuals are not normally distributed.

Issue with this is that such a question is inherently more noisy than a latent factor, like g vs. one specific item, so all correlations will fall, including the heritability and the correlation with paternal age. 

This is a questionnaire frequently used so it serves as a good reference point. If different research teams use different measures, their measures lack comparability. You may argue that your measure is better but it should be made clear(er) in future studies.

1. First of all, I have to admit I missed an important detail earlier: the first figure PCA Scree Plot did not have a title (about this one, if you could increase its resolution, this would be excellent). This should have been your Figure 1 (which is also not mentioned or discussed in the main text) so your Figure 2 and following ones should all be re-numbered as well as in the main text.’

Done

2. It should be: mean = 61.3, SD = 7 years

Fixed

3. "both" factor analysis?

Fixed

4. In your abstract, d is italicized but not here.

Fixed

5. When presenting Figures/Tables in the main text, the typical presentation is that they are called out (refered to) in the main text before, not after the Figure/Table placement. I'm not too picky on this, I'm fine with both, but in your situation, Figure 3 is called out before, while Figure 1-2 are called out after. At least be consistent for all tables/figures. 

Fixed

6. Actually, figure 4 shows only r, CI, and n, but not the p value. Since you wrote that the correlation is "statistically significant" it is odd not to disclose the p value as well.

Obvious based on the CI but I added p < 0.001 in

7. There should be a dot after Figure 5 (same issue for Table 1). I ask that you reread the paper very carefully, because there were tons of mistakes like these ones. Perhaps I even missed a few more. Also, the title of Figure 5 is not italicized but all other figures/tables are. I don't think the OP journal requires you to italicize the Figure/Table title, and usually don't see these in italics in other journals. So this is very odd. Finally, another nitpick with regard to Figure 5: I think it's better you change the title as follows, e.g.,: Correlation matrix of study variables" because everything is odd here. 

Fixed

8. Same issue as above; between study variables looks much better.

Fixed

9. Since you're not using APA citing style, this should be Bronski [1].

Fixed

10. Spacing.

Fixed

11. This is another nitpick. After given some thoughts I believe it's best to use formal words for professional papers. So "nice" is a bit odd here. 

Changed to a bigger word

12. That was still a weird wording for me, because with one factor solution, why rotating?

I could be wrong here as I’m only just reading about this, but varimax is done after PCA. So the PCA version is just the unrotated factor and varimax is applying rotation to that. PCA automatically gives you a number of components equal to your columns. Varimax rotates more the more the PC loadings on columns cluster. https://www.youtube.com/watch?v=AjrU9oV3MRM

So probably what is happening here is varimax is not rotating much, but my items are probably all slightly loaded on the second PC as well as the first, so this is rotated to make them only along 1 dimension. So the rotated solution is better if I am only taking the top factor.

13. Fine, but at least make it clear in the text. Typically, researchers don't use robust SE unless they suspect or know the residuals are not normally distributed.

Added a sentence on this

Author has updated the submission to version #3

After reading the new version, I accept the publication.

Just make sure you present the results from Figure 1 in the main text, as you did for the others.

The author continues his investigation into whether de novo mutations and thus paternal age affects political views. The author presents two pieces of evidence consistent with this hypothesis: 1) paternal age correlates with offspring leftism and 2) paternal age doesn’t correlate with parental leftism, implying result 1 is not due to genetic confounding. Although the author has shown result 1 in a prior paper, the use of a continuous measurement and a larger sample size creates much stronger evidence for paternal age correlating with leftism. 

I’m sceptical of the author’s conclusion, nonetheless I greatly enjoyed the paper and think it makes an extremely important contribution. With some changes I think the author can make his case more air-tight. Nevertheless, the paternal age effect on leftism is now reasonably established and begs for an explanation. 

I number my advice to the author below:

1. Could the author provide some background explanation for why de novo mutations causing leftism is plausible? The idea to most readers will seem outlandish. Some explanation (e.g. de novo mutations have been shown to cause many traits correlated with leftism) would make the manuscript much more convincing and compelling for readers to take seriously.

2. I appreciate the author’s brevity, nevertheless would he expand on possible alternative explanations for the paternal age effect. Could there be a sociological explanation? Could one’s upbringing be different with an older father.
3. The author suggests birth order effects do little to confound paternal age correlations for IQ and mental illness and thus suggests “It is unlikely birth order explains much of the paternal age correlation.” I find this to be a weak inference to make, which I might qualify or better justify with more research comparing birth order effects and birth order
4. Result 2 does not clearly disprove a large role for genetic confounding. In Figure 8 there is a confidence interval is between -0.14 and 0.09. Now these confidence are overly small because the author has not accounted for the fact that the observations are not independent - multiple come from the same father. I would encourage the author to rerun the correlation using clustered standard errors or bootstrapping confidence intervals by randomly choosing fathers to include in the data.

How large would the correlation have to be for the paternal age effect to be entirely genetically confounded? Let us assume an additive heritability of politics of 50% and perfect assortative mating (outlandish, but the truth is not far off), then the correlation between father’s age and child’s leftism is the correlation between father’s politics and paternal age multiplied by the heritability. Under these assumptions we only need a 0.2 correlation between father’s politics and paternal age to explain a 0.1 correlation between paternal age and leftism. In other words, with larger confidence intervals a non-negligible role for genetic confounding may not be ruled out. The author amy wish to provide such a back of the envelope calculation (with less extreme numbers) to make clear how convincing his results are.

With correct confidence intervals, I would then caveat that the results do not make genetic confounding impossible, but the results make it of much less significance and much less likely to account for the paternal age effect. 

5. I think result 2 would be more convincing if the authors also found a much larger publicly available dataset to perform this analysis in. The GSS has the variable kdyrbrn giving the year in which someone had their nth child. From this you can calculate the correlation between leftism and parental age. It has an N of 1126, but might just include mothers. Again, I would encourage bootstrapping observations by parent or clustering the standard errors at the parent level
6. Table 1 must be changed from console output to a proper regression table. Console output is considered extremely unprofessional. 

In summary, I demand the reviewer makes a proper regression table and clusters the standard error and confidence interval for result 2. I encourage the author to replicate result 2 in another dataset (e.g. the GSS) and describe more the motivation for his theory and more seriously consider alternative interpretations. This will make his paper and theory much more compelling. 

I like this paper and look forward to seeing it published. Below I point some issues with typos and presentations which should be cleaned up.

This measurement has high reliability (Cronbach’s alpha = 0.93) which far outperforms the commonly used Wilson-Patterson Conservatism Scale (alpha = 0.71)

Be sure to keep the reporting of cronbach’s alpha consistent throughout the text. APA format is to use the greek letter inside the brackets. I think this is the most beautiful.

as well as high validity (leftism d for Republicans and Democrats was 2.31, p < 0.001)

This doesn’t look very pretty in an abstract. Maybe try something like “as well as high criterion validity, as evidenced by the Cohen’s d between Republicans and Democrat on the measurement (d = 2.31, p < 0.001)”

The first surveys 1175 white American men, mean age 41.5 years (SD = 13.2 years) and gives them the general leftism test, and asks their father’s age when they were born (mean = 61.3, SD = 7 years).

No need to put years inside the bracket

Also, items F2, F3, R2, and R3 were reversed. 

This probably shouldn’t be its own paragraph

We achieved factor loadings of 0.88, 0.87, and 0.78 for race, feminism, and gay respectively. 

I think referring to a facet as gay sounds a little weird. Perhaps refer to it as the LGBT facet. E.g for race, feminism and LGBT facets respectively. Generally I think LGBT sounds a little more professional and scientific whereas gay could sound unnecessarily pejorative. 

Cronbach’s alpha for the three sums was 0.86, which is far over the typical significance threshold of 0.70

This is not a significance threshold but rather a heuristic for if a measure has good internal-consistency reliability.

p<0.001

Missing space between p and inequality sign

Author has updated the submission to version #4

I have fixed the typos and done everything else you asked except for updating the tables. These I'll do last, because I don't want to have to edit styled tables but once. So if you have no other criticisms, I will do the tables but otherwise I will do them later when it is sure that the tables will not be changed any further.

Importantly, I have expanded my design for study 2 because I did not feel that the original sample size was adequate once you pointed out the need for clustered SEs. I have also decreased the certainty level of my verbal conclusions because of this. I think it is unlikely, but I cannot rule out older parents being more leftist as a confounder. In the new data, I ran out of men so I had to turn to women. Because of this, I controlled for sex as they are slightly more leftist on average. I have not updated study 1. 

I have also added a model and a more detailed discussion of the molecular genetics of mutational load. I call for a larger, superior study to this that probably can't be done on prolific that would directly control for parental leftism as well as birth order, maternal age, and many other potential confounders.

I hadn't replied since I didn't want to give more comments until all edits were made. But the review editors has asked for some comments. In general I would create the tables before asking a reviewer for comment, since the reviewer may have important qualifications to make on the table itself. It’s a real pain to recreate the tables after each change, but at least in R there are some decent packages that can output the tables in a format you can copy into word or latex. 

Expansion to study 2

I appreciate the nuanced discussion of the results. I understand you have increased the sample size. I think using a larger publicly available dataset would have been interesting, but now I suspect that's an analysis for another paper. 

I don't understand the logic of the new control for current age. After controlling for current age, age at birth rearing just tells us the effect of having kids so many years before the present, not the true association between parental age and parental leftism. I would not control for this variable. If you think you should control it then explain the choice. If you think it’s ambiguous then report estimates with and without the control

Using mothers is interesting, since we really expect the mutations to come from the men; it is primarily the men for whom we are worried about being confounded. So it is their leftism-rearing age correlation which really matters. But I agree using mothers is useful to increase the power and the correlation will not be too different. It is also somewhat relevant since we expect father’s age to correlate with mother’s which may also induce confounding. Nevertheless I think you should report the result with just the men. You could do this with an interaction term between age and sex. You can still report the combined sample as well.

The Mutation Model

The section genetic model of intelligence decay isn't very clear. I appreciate the effort to provide a well theorised expectation for there being a paternal age effect, but there are a few issues:

I'm not familiar with the mathematics of mutations, but the mathematics and equations appear incomplete. No reference is made to the plausible effect size of a mutation on leftism, or the heritability of the trait more generally. The first equation formulates the paternal age correlation in terms of number of genes, number of leftism genes, mutations per generation and a few constants. I can't see how the equation could be correct since it does not incorporate the actual effect of the mutations. In a world where the heritability is 1% versus 99% the effects of mutation should be very different. 

Maybe the constant 2/3 is determined by the effect size of the genes. But this is not explained nor justified. If possible a source should be given for the equation. If there is none, make clearer which sources you have derived it from.

Give a source for the claim the human genome is 2mb in size.

The subtitle should not be "Genetic model of intelligence decay" since the paper is not about intelligence. I think a genetic model of the paternal age correlation, or something similar would be more apt.

There are many issues with compiling the math ie. sometimes delta is given as the word and sometimes as the symbol, subscripts are sometimes compiled, sometimes left as "n_t" for example. 

In general this whole section is poorly cited and explained. It may be worth ditching it entirely. The math could also be left in a supplement or an appendix. In general given the challenge of doing this sort of analysis convincingly and correctly it might be worth leaving it for a whole separate paper on the topic.

Hi Bronski, It's been two months since my reply and three since your last update. I'm just writing to remind you to turn in a new version when you can. I'm anxious to approve the paper quickly. 

It has been half a year since your last update. This is just a reminder so we can get the paper published promptly.

Author has updated the submission to version #5

Updates:

1. Styled the tables

2. Modified the math model to include heritability

3. Fixed typos 

4. Changed model name

5. Specified there was no interaction between sex and breeding age. 

Author has updated the submission to version #6

Further typo cleaning and added a couple of newer references.

I appreciate the clarity in your equations and it looks like the estimated range checks out given the model's assumption. 

A few things (minor details) you might want to consider.

Make sure the first and second paragraphs of the method section are separated by spacing them.

For Table 2, I think you might also want to report the unstandardized betas for all predictors (either added in the table or mentioned in the main text). Since sex is a dichotomy, it's more sensible and interpretable in unstandardized beta. It may be hard somewhat to interpret the meaning of a one standard deviation in sex. 
I still don't like the terminology of "partial correlation" since it's more accurate to describe it as regression coefficient. But I guess I'll let it go.

Since a relation between paternal age and general leftism of just 0.055 cannot be ruled out given the confidence interval from Table 1

I remember the confidence intervals were there in the earlier tables, but it's no longer here. Mayb add the confidence intervals back directly in the table. Also, it seems this value of .055 refers to table 1 as well? It doesn't correspond to anything I'm seeing.

.11 - 1.96*.28, the lower part of the 95% CI.

I'll add a paragraph break as you requested, add explicit 95% CIs to the regression tables, and as for standardized vs. unstandardized betas, I think it would be best to add a table with all the variables' means and SDs. Unstandardized models can be recovered from this combined with the standardized models.

Sorry for my late review. 

I would strongly suggest removing the mathematical modelling. I've spent more time thinking about the model and I'm not convinced it is good model, and I think there may be at least one mathematical mistake. The effect on the phenotype should be h, not h2, times the change in the genetic score. h is the correlation or effect of the genetic score on the phenotype. You might have thought h2 is correct because you are familiar with the breeder's equation. The reason why the breeder's equation uses h2 is because it involves the effect of selecting the parent's phenotype on the genetic score and then the effect of genetic score on the child's phenotype. Here the correlation between phenotype and genetic score occurs twice! 

Below I list a number of (often implicit) assumptions in your model.

- seems to assume all mutations come from dad

- Humans are haploid with 0 or 1 copies of any allele at 30,000 different loci (the 30,000 loci is based on an idea that each gene should have only one associated loci). 

- There are 40 mutations per generation (Current estimates I think can be very different and also very uncertain)

- one-to-two-thirds of alleles have a 0 effect whilst the rest have an equal effect on IQ. 

- all allele frequencies are 0.5 to start with (and the risk of mutations hitting twice at the same allele are ignored).

- The effect of paternal age is 1/3 of the average effect of mutations per generation (a blog post is cited for this assumption. After a very quick scan I could not see the derivation). 

Of course these assumptions are unrealistic. Models are meant to be simplifications to illustrate certain facts. The critical issue is that your choice of model assumptions determines the effect of mutations on IQ. This is what we need as a benchmark for how mutations might affect politics IF politics is like IQ with all mutations affecting the trait in one direction. The estimated effect of mutations is only as compelling as the assumptions that created that estimate. The assumptions are not realistic enough to be compelling.

Even if the h2 error was altered, I could not accept this model being published. The assumptions are not compelling and I worry that I lack the time and ability to be sure there are no other errors. You have a number of options 1) remove the model. 2) remove the model from the main paper, fix the h2 error, make it supplementary material and suggest that is speculative. 3) You come up with an entirely new model that is compelling and if the model is too complex for me, have the submission editor find a statistical genetics to review it. Given how long the paper has been in review and how difficult it is to come up with a compelling model I recommend 1). 

If the mathematical model is removed I'm happy to accept publication. The regressions are interesting. 

A few minor things. I agree with reviewer 1 that regression betas should not be called partial correlations. 

Standard errors are robust to cluster correlation, as participants had multiple breeding ages each.

This is written unclearly. Best to just say that cluster robust standard errors are used.