Thank you for the statistical analysis. It does shed light on FRC’s claims.

And your hypothesis does offer a possible explanation (assuming that the whole thing is not bogus – we won’t know until we find out why 1/4 of women are not accounted for). I’d also like to have you consider an alternate hypothesis, using the same logic.

If it can be shown that women from cohabiting step families (CHS) have statistically higher incidences of same-sex expression, this is only relevant if it is uniquely same-sex.

In other words, it may simply be that CHS women are having more sex in general than those from an intact family structure (IFS). If IFS women are not having a lot of sex, then it would be only expected that they would have lower levels of same-sex activity – along with lower levels of opposite sex activity.

I think it likely that if we ran the numbers for opposite sex activity in the last year, it would likely “prove” that women from intact families are both less homosexual AND less heterosexual.

Which, of course, is nonsense.

However, if you remember, Professor Sullins said we needed more advanced statistical procedures to know anything with more certainty. Well, the ANOVA is a far more accurate way of determining significance when you have multiple groups that by performing individual t-tests. When you do individual t-tests for all the conditions involved, you are drastically increasing the experimentwise alpha level (the possibility of a type I error). So, in order to have a more certain result at the same confidence that one individual t-test would provide, we should test the groups all at the same time using the ANOVA. I did that last night and the results are clear: The only significant difference was found between the cohabitating stepfamily and the two intact family structures. All other relationships are non-significant. To put it another way would be to say that married intact, cohabitating intact, always single, married stepparent, and single-divorced all had about the same mean average number of women engaging in same-sex behavior. ANOVA is the more accurate statistical tool and this is the conclusion it shows. If you would like the graphs, I’d be happy to send them to you, but I repeat my assertion that the FRC study has not shown anything like they purport. The truth is that the difference between the average amount of same-sex behavior among women originating from non-intact families of origin IS NOT significantly different from the average amount of same-sex behavior among women originating from intact families of origin except in the case of cohabitating step families. And I believe my hypothesis from my previous post lends a possible explanation for why this is true.

Explain why you would want to test for homoscedasticity with this dataset. Get as technical as you want!

1 2 3 4 5
0 NS NS .003 .015 NS

1 NS NS NS NS

2 .014 NS NS

3 NS NS

4 NS

This is huge. It shows there is only a significant difference between the married intact group and both the cohabitating step family and the divorced single family. That means that there is no difference between the amount of female-same sex behavior in married intact groups and both the married stepparent and always single parent group. The only other significant difference was between the intact cohabitating group and the cohabitating stepfamily.

I also ran an ANOVA with the numbers and it shows no significance between any groups except the intact marriage group and the cohabitating stepfamily group (which also was significantly different from the intact cohabitating group).

What does this mean?

1. It means that there is no more female same-sex behavior among women who came from always single-parent homes and married stepparent homes than from intact married homes. This is huge. And there’s no doubt that if this is true with the sexual behavior that no correlation will exist among the 1.8% of lesbians. It jives with some research I read through Warren Throckmorton’s site last night – namely that family structure does not influence sexual orienation.

2. If the results of the ANOVA hold up it will show that the only differences are between the two intact groups and the cohabitating stepfamily. My theory would explain that very well.

So, once again another FRC smear job bites the dust. :)

(Tim, if the chart doesn’t display correctly, I’d be happy to email you my chart as well as all the charts from SPSS that show the calculations and which ones were significant)

Great report. I’m glad to know that it wasn’t my researching abilities being poor that kept me from finding the data LOL. I’m glad that Professor Sullins went beyond the chi-square test, because they have a lot of built-in assumptions which is why they are not used very often. I agree with his analysis. If the t-test shows significance then it is a significant finding because the sample size is not trivial, though I would like to know, along with Tim, why the numbers are different. This may seem picky, but FRC wants to be taken seriously and yet they report the wrong numbers in their ‘study.’ That’s totally unacceptable. Again, Sullins is right about needing more advanced statistical tests. Let me suggest a few things from a further review of the data I did tonight and end with a hypothesis that would explain the significant findings:

1. Something I said in my last post and something Tim pointed out in the main article is that the relationship here is between sexual behavior and family structure. Even if the relationship bears itself out in further analyses, it proves nothing about orientation. This is critical! Anyone who has done even a little reading on the topic knows that human beings engage in sexual behavior for all sorts of reasons besides in fulfillment of their sexual orientation. For example, in this study, overall, 11.2% of women reported a same-sex sexual experience in their lifetimes, but only 1.8% identify as lesbians. That means of all women who ever engaged in same-sex behavior, only 16% of them identify as lesbians. So, right from the outset, we know this study doesn’t prove anything about sexual orientation per se since the overwhelming majority of those engaging in these behaviors tell us they’re straight!

2. In their book “Born Gay”(2008), Glenn Wilson and Qazi Rahman review the most recent studies on sexual orientation identification. It seems that, despite the common assumption of a large number of bisexuals, they actually only make up about 0.2% to 1.0% of the population at large, perhaps a bit higher depending on how the question is asked. In this study, the overall number of women identifying as bisexual is 2.8% which is significantly higher. However, if you break it down by age group an interesting pattern develops. Here it is:

18-19 = 7.4%
20-24 = 3.5%
25-29 = 2.8%
30-34 = 2.1%
35-44 = 2.0%

Notice the decline. The number of those identifying as bisexuals goes down 73% over the course of 25 years or so. What should our conclusion be? Wilson and Rahman (2008) observed the same pattern in their research and I just read a couple journal articles this week that identified the same trend in their data. It appears that the bisexual categories become depleted over time as people become more experienced and discover who they are with more certainty (and at that point identifying as hetero or homosexuals). And, as you can see, the change between 18 and 29 is far more dramatic than the change between 30 and 44. This conforms to the aforementioned research findings which all indicate that by the mid to late twenties we stop seeing the shift in identification as people consolidate their identity. Just to reinforce this and to lead into my next point I will show the same breakdown for women who have had a same-sex sexual experience in the past twelve months.

15-19 = 7.7%
20-24 = 5.8%
25-29 = 3.6%
30-34 = 3.0%
35-39 = 4.5%
40-44 = 2.4%

You see the same steady decline, suggesting that fewer women are having sexual experiences with women because more of them are consolidating their identity (the only anomaly was the 35-39 year old age range, but their numbers were different in all the categories across the table, so there may have been a sampling issue there or perhaps women get hornier in that age range LOL J/K). And it needs to be pointed out that between 25 and 44 the percentages are BELOW the overall 4.9% average. So, it appears that the most active women are those who are younger, again confirming my earlier thoughts.

3. All this leads me to a hypothesis that may explain the correlation. We know that people who emerge from more unstable homes tend to have more emotional baggage than those who emerge from intact homes. People from intact homes tend to have a more solid sense of who they are, because their developmental path through a stable home environment allowed them to consolidate that identity (including the question of their sexual orientation) during their teen years (thus, not as great of a need to sexually experiment). So, my theory stands to reason, that those from non-intact homes, with greater unresolved emotional issues have yet to consolidate their identity and, as such, have not yet identified their sexual orientation with the certainty that those from intact homes have by the time they reach high school graduation. The delay in identity formation means they would continue to engage in same-sex experimentation as part of the self-discovery process. In these cases where the women have not completed that developmental task in high school or early college, they tend to complete it sometime in their twenties, at which time we would expect the same-sex sexual experimentation to decline and this data clearly suggests that it does. THEREFORE, if the data that Professor Sullins provided can be broken up into age ranges for each of those family structures and the numbers decline in the same way as the previous examples, then the variable of “delay in identity formation process due to emotionally volatile home life” could potentially be identified as the causal factor in the statistically significant differences FRC put out and a demonstration of those numbers decreasing as we go up through the age ranges would show that the process of consolidating one’s sexual identity is still going on in those groups, albeit delayed due to the home environment. THUS, the conclusion would not be that non-intact families produce lesbianism as they wanted to spin it, but that non-intact homes cause a delay in the developmental process of an adolescent’s consolidating his personal and sexual identities, leading to higher levels of same-sex experimentation, but that they catch up with the process by the end of their twenties and the same-sex experimentation levels drop off.

If FRC wanted to show this correlation with the 1.8% of self-identified lesbians, they could have made their case that family structure makes one gay. However, it is my suspicion that the data does not show that, so they went with the next best thing. As far as the religious services thing is concerned, that’s simple. Those who grow up in non-intact homes attend religious services far less often in general. We know that already from other research. Thus, it’s not surprising that the women who engaged in same-sex activity who came from those homes would be less likely to attend services as regularly anyway. That’s a no brainer. And it’s also insulting because the implication is that those who don’t go to services regularly are immoral people and “this proves it!” Circular reasoning has always annoyed me. Peter, Tim, I’d like to hear your thoughts on this when you get some time.

I looked over your info and the first things I noticed were:

1. Your chart doesn’t add up.

2. The total women in this analysis are about 2,000 fewer than the total of all women in the study.

3. The FRC average for all women was 4.9% but the CDC reported an average of 4.4%. I’ll have to look further to see why, but the sample size for the CDC seems to be larger. I was under the impression that it was the same database, but I may be mistaken.

Bottom line – on the family background / lesbian activity data the differences are statistically significant for all comparisons to “intact with married biological parents”. I’ll post more in my blog when I have a spare moment.

I’m not sure it’s valid at this point to say “it’s likely an illusory correlation” (though I accept that’s just a supposition). What might be useful is putting together some testable hypotheses on what the correlation is about and then seeing if we can go about exploring them.