Gifted Issues Discussion homepage Forums Parenting and Advocacy How to discuss gender inequalities

If you calculated your inter-country variance based on the information from Table 2, your data set is incomplete. The purpose of Table 2 is to demonstrate whether consistency exists among different tests given at different times within the same country. If a country did not participate in enough different testing cycles to shed any light on this, its data was left out of Table 2.

When the authors then comment on the "large variance" you're taking issue with, they do so like this (with figure 1A being the histogram in question):


We can look at the 2007 TIMSS 8th-grade column on Table 2 and find Tunisia on the low end with 0.91. We cannot find a high of 1.52... the highest figure provided is 1.31 for Taiwan. The study states that 8th graders in 52 countries participated in the 2007 TIMSS, yet exactly half of those results are reported in Table 2.

I'd agree with the statement that 0.91 to 1.31 is a wide variance, and 0.91 to 1.52 even more so.


Since the study is out to determine how mathematical performance differs among the sexes amid different cultures, that was the right choice.

Some more interesting data which came from other studies and were mentioned in the intro to this one:



So, what biological revolution occurred in the US between the sexes from the 70s to today that can explain these results? Answer: none. What social revolution can explain them? Answer: feminism.

*looks around for the "like" button*

Omg lol. I told the hubby about what Wyatt said about "girls can't be astronauts" but then when we watched superman he said "the girl is probably a secret superhero too because she has gloves and girl superheroes wear gloves". The hubby said, "so girls can be sidekicks!"

The reference is "Sky High" a superhero movie where kids go to superhero school and are divided into hero class and sidekick class depending on how cool their powers are. Of course all four of us, counting the baby, call each other "sidekick" since the movie. Of course I'm really the superhero with three sidekicks even though these other three say the same thing.

Table 2 is restricted to information from countries in which 3 separate measurements had been made. By averaging these samples, the result should be more reliable than those from countries in which fewer measurements were taken. The selection process had nothing to do with their statistical consistency, so any bias is unintentional. If you would like to compute a mean and variance for all the inter-country VR measurements available, please be my guest.


My point is that it's sloppy (or misleading) when you are presenting a statistical analysis to quote the range of variation as being large without calculating and stating the variance. The range is largely meaningless, while the variance is telling.



The authors take their performance data and use it to comment on biology. Their tests are not well suited for this purpose.



When you realize that we agree on this you'll be one step closer to understanding my position.

Similarly, this

is a complete misrepresentation of the natural greater male variance explanation, based on the assumption that this explanation excludes the possibility of any non-biological effect on performance, as well as the assumption that if the explanation were true for any group it must be true for all groups around the globe. I don't agree with these assumptions.

In this way, the authors essentially argue that the evidence they have purportedly demonstrated in support of a nurturing effect absolutely contradicts any natural discrepancy. However, it only contradicts a nature-ONLY discrepancy. Just because nurture is effective, doesn't mean biology plays no part.

In order to determine what the natural VR ratio should be, we have to try and strip away or control for the other contributing factors to that measurement. This study did not do that. The authors demonstrated that VR varied across populations, and jumped to the conclusion that the natural VR simply must be 1.0. Maybe it is, but that has not been demonstrated by the study. To do that is a much more ambitious undertaking, with difficulties I alluded to in earlier posts.

Then just say that, instead of substituting the missing value with a sloppy and misleading calculation of your own. There were only 52 values to start with, and you can't non-randomly reject half of the sample size and pretend that the result has any meaning.

The study did cite their source, with the expectation that anyone wishing to check their work would use that, rather than Table 2.


I understand your position, thanks anyway.


The authors say nothing about what a natural VR should be, other than to say that if the greater male variance hypothesis is true, then it should be demonstrable throughout wide cross-sections of the human genome... and national cross-sections will do very nicely, because they encapsulate a vast enough sample size that any individual nurturing/opportunity gaps should cancel out, leaving us a data set that is normed for the cultural and biological group.

Then they go on to observe that under improved gender equality conditions, VR approaches 1. This says when cultural conditions are equal, so are performance results. If there were a biological component involved, this should not be true... some inequality in results should remain despite equal opportunity. Therefore, the greater male variance hypothesis is false.

I do agree that their statement about VR being "similar in value" across nations does take things too far, and does argue for a "biology only" position. Some fluctuations due to social influences should be present. But if gender equality means that VR approaches unity, then that is a powerful "not biology at all" argument. And that's what the data shows.

I take this to mean that the barriers to including all the values are too great for you to bother with as well. Also, you absolutely can throw out values that may be less reliable for those that are believed to be more reliable. The result is meaningful. If you want a different calculation for comparison, go make it.


What section are you referring to here?


Seriously, are you just making that up or is my copy of the study missing pages?

The countries the authors call out as having nearly equal VRs include the Czech Republic (GGI=0.6718), Indonesia (GGI=0.6550), Morocco (GGI=0.5676), Tunisia (GGI=0.6283), and the Netherlands (GGI=0.7383). Outside of the Netherlands, those are low GGIs (sometimes very low). I think that's why they specifically stated


rather than stating as you have implied that countries with more gender equality are the ones with VRs closer to 1.

You don't have any reason to believe those values may be less reliable. The same testing method has already proven itself to provide consistent, reproducible results among 26 other nations.


That was a read-fail on my part. I retract it. I read examples like this and interpreted "gender gap" as GGI, where what they meant was a gap in performance:

"For example, they were essentially coincident in some
countries, such as the Czech Republic, where VR
and gender gap were near unity and zero, respectively
(Figure 2A)."

So then, will you agree that this study constitutes less than a thorough debunking of the greater male variance hypothesis, despite the authors lofty claims?

I guess it's your turn again for a read-fail, because the authors make no such claim.

There is no support in their study for the idea that greater male variability is "largely an artifact of sociocultural factors". For all we know, it's largely a result of biology, with sociocultural factors (along with sampling problems) causing the tails of the VR distribution to vary from a mean that is greater than 1.0. As you have stated, the author's did not estimate what a natural VR would be. They did not perform any regression analysis to determine what factors were at play in the VR measurements they obtained. They simply skipped all that and labeled this as "largely an artifact of sociocultural factors". It's an unsupported assertion. Thus my "lofty" claims objection.