As before, I think you should use as predictors the national K numbers in Meisenberg & Woodley, 2013. Other than that, this is a good study.

This is a good paper, but I would like to see the author clarify one point. Just what relationship is being postulated between criminality and belief in Islam? Is the author suggesting a direct relationship? In other words, belief in Islam, even among converts, leads to increased criminality. Or is the relationship more distal? In other words, most Muslims lived until recently in pastoral societies with weak State formation. Every adult male was expected to use violence pro-actively to defend himself and his family from real or perceived threats. Gene-culture co-evolution thus favoured lower thresholds for expression of anger and greater pleasure from ideation of violence.

I have seen no studies on the relative criminality of practising Muslims and nonpractising Muslims, but the perception, particularly in France, is that criminality is much higher among assimilated secularized Muslims of the 2nd and 2rd generations.

You wrote: "Recently I analyzed crime and fertility among immigrant groups in Denmark by their country of origin[1]. I found that they were highly predictable using information about their countries of origin using only 3 predictors..."

This is incoherent. Try:

"Recently, I decomposed the crime and fertility rates of Danish immigrants by country of origin[1]. I found that these rates could be predicted with a high degree of accuracy by three nation of origin variables: Islamic belief, average height and either national IQ or GDP. Other predictors like country of origin murder rates were not good predictors. Here I extend this analysis using two datasets from Norway. The first data set is from the social statistics agency (Statistisk Sentralbyr a). The second is from a 2011 report, "Criminality and Punishment Among Immigrants and the Rest of the Population (Kriminalitet og strablant innvandrere og vrig befolkning)" [2], which was released by the social statistics agency. "

You wrote: "Initial results showed that data for Mongolia was an extreme outlier and it was therefore deleted. Most likely due to data error."

Data error or sampling error?

You wrote: "There are a number of reasons to believe the dataset to have some systematic error that reduces all the correlations with predictors."

Could you explain why this is surprising? What, for example, was the NIQ-National crime rate correlations. And why aren't you using national crime rates as a predictor?

You wrote: "Third, calculating the correlations between the predictors (IQ, GDP, Islam) and crime rates for the individual years shows that all 11 correlations are negative for r (IQ x crime), all 11 are negative for r (GDP x crime) and 10 of 11 are positive for r (Islam x crime). These results are very unlikely if crime rates were not predictable, but expected if correlations are systematically reduced."

You could just compute the cross year joint probability.

You wrote: "Fourth, the extreme differences between immigrant groups are not believable. The most criminal group is Georgia with a rate of 79 charges per 100 persons (!).

Ok, are you weighting these groups by e.g., sample size or SQRT of population size?

You said: " I contacted the author of the report to hear if there were more information available, especially for a larger country sample set. He told me that this was not the case and that one needs to request speci_c data from the statistics agency to acquire the data. If it is like in Denmark, this can be quite costly."

Get rid of this nonsense.

You said: " Next up..."

Rewrite.

"I think some kind of causal connection is plausible. The correlations are very high, also for the Danish data and not did become small even after using partial correlations with IQ, GDP, height as controlled. It is possible, of course, that the correlation is due to some confound, but we just haven't found it."

I believe we agree that there is a causal connection, but we differ on the nature of the connection. Is it direct or indirect? By using terms like "Islam belief", "prevalence of Islam" or simply "Islam", you create the impression that the causal connection is direct, i.e., if you control for genetic and cultural factors, belief in Islam will be associated with a greater predisposition to criminality.

This causal connection has not been demonstrated. In fact, non-practising Muslims seem to be more crime-prone than practising Muslims. In your present study, no distinction is made between the two groups. All immigrants from Muslim countries are coded as having "Islam belief".

Many Americans will similarly argue that Black on White violence is ideologically motivated, i.e., it's due to "anti-White racism." A more parsimonious explanation is that such violence is largely a spillover from the high rate of violence within the African American community. Yes, there are many cases where young Black men deliberately target Whites, but this is not so much anti-White racism as the universal tendency to seek out "soft targets." In other words, "Whites don't fight back." In general, people in European societies are not accustomed to fighting back on their own, preferring instead to put their trust in State institutions for protection. This is their Achilles heel, and many people out there are gleefully exploiting it.

This is admittedly a complex debate with many pros and cons. I don't expect you to cover it in a short paper. But please don't assume what you wish to prove. For this reason, I recommend that you use a more neutral term like "Muslim origin" instead of "Islam belief".

The weighted results from dataset1 (the damaged one).

I'll ask a statistician friend about the weighting issue. In the meantime, I noticed that your crime variables were highly positively skewed. (Sorry!) Given this, you have to either:

( a) Transform your data e.g., log transform as shown in the example
( b) Use non parametric statistics
( c) Use robust statistics

Try (a). See example.