# HOMOSEXUALITY IS NOT HEREDITARY

Two recently published "scientific" reports claimed that homosexuality is hereditary. The press has taken this conclusion as proof that the homosexual person has no control over his or her sexual preference. This is then touted as a reason homosexuality should be a protected civil right. But a detailed analysis of what was actually done to make these reports shows that this conclusion is not only unwarranted, but is in fact proven false by their own data.

## STUDY 1

Note: A link to the website of this study used to be here, but the site no longer exists.

According to the first published report, the scientists tested 15 subjects known to be homosexual for the presence of a certain gene. 4 of the 15 individuals tested possessed the gene. They reported that a test using Student's t distribution showed that 26 percent is a statistically significant value, and concluded that homosexuality is hereditary.

Now, let's look at their methods and mistakes:

1. No control subjects were included in the experiments. All of the subjects studied had the wanted trait (homosexuality). No control group was present. Therefore, there was no independent variable in the experiment -- nothing to compare the result to. You can't even prove a correlation, let alone a causal relationship, without a control group to compare findings with.
2. The experiment was designed wrong. The gene should have been the independent variable, and homosexuality should have been the dependent variable. Both values should have been collected over a large number of randomly selected subjects.
3. The collected data were categorical data, not numerical data. But they used methods for numerical data.
4. The sample size was too small. A good study needs at least 30 subjects, either randomly picked, or 15 in the experimental group and 15 in the control group. A larger number of subjects increases the significance of the findings. Using categorical data requires even larger sample sizes. And a larger sample size is needed if the value tested for is relatively rare.
5. They got the math wrong. 4/15 is .2666 ... which rounds to 27 percent, not 26 percent.
6. They used Student's t distribution, which is used to compare two experimental values (numeric), or to compare a value to a norm. But they had neither two experimental values, nor did they disclose a norm to compare their one experimental count to. This test is designed to be used when the data are multiple values of some real-number (numeric) variable.

But the data at hand were discrete categorical values (Boolean type yes-no values). The correct method is to create a crosstable, calculate the Chi-squared value, and then perform a Chi-Squared test to determine whether or not there is an association. But they didn't collect the data needed to do that.

7. They did not reveal whether they tested for the presence of the DNA itself, or for a protein that is produced only if the gene is active.
8. They released the results as though they had shown a causal relationship, without any proof of a causal connection, or even an association. Since the data were categorical, a correlation is impossible.

## STUDY 2

Note: A link to the website of this study used to be here, but the site no longer exists.

According to the second published report, the scientists tested 40 subjects known to be homosexual for the presence of a certain gene. 33 of the 40 individuals tested possessed the gene. They reported that a test for correlation (to what???) showed that 64 percent is a statistically significant value, and concluded that homosexuality is hereditary.

Now, let's look at the methods and mistakes in this second study:

1. The method used to obtain subjects is highly suspect. They advertised for volunteers in homosexual-interest magazines.
2. No control subjects were included in the experiments again. They don't learn! All of the subjects studied had the wanted trait (homosexuality). No control group was present. Therefore, there was no independent variable in the experiment -- again, nothing to compare the result to. You can't even prove a correlation, let alone a causal relationship, without a control group to compare findings with.
3. The experiment was again designed wrong. Again, the gene should have been the independent variable, and homosexuality should have been the dependent variable. Both values should have been collected randomly from the general population.
4. The collected data were categorical data, not numerical data. But they used methods for numerical data.
5. The sample size was still too small for any great level of significance.
6. They also got the math wrong. Where in the world did they get 64 percent? 33/40 is .825, which rounds to 83 percent, not 64 percent. If they took the distance of .825 from .5, they would have gotten 65 percent, not 64 percent. Another quandary!
7. They used a correlation test. But who knows WHAT they correlated the data to? Again, they had neither two experimental numeric values, nor did they disclose a norm to compare their one experimental count to. This test is designed to be used when the data are multiple values of some real-number (numeric) variable.

But the data at hand were discrete categorical values (Boolean type yes-no values). The correct method is to create a crosstable, calculate the Chi-squared value, and then perform a Chi-Squared test to determine whether or not there is an association. But they didn't collect the data needed to do that.

8. They did reveal that they tested for the presence of the DNA itself. This means that the distribution is the same as that for a dominant gene. But their statistical test didn't seem to take this into account.
9. They released the results as though they had shown a causal relationship, without any proof of a causal connection, or even an association. Since the data were categorical, a correlation is impossible.
10. An attempt to duplicate the second study was unable to reproduce the results.

## Analysis of the genetics behind the studies:

Let us see if we can figure out what norms could have been used in either of these studies:

• If the expected norm was the actual prevalence of this DNA in the population, its value was not disclosed.
• If the gene is randomly distributed in the population, it should appear in roughly half of the DNA passed to offspring. But we have to remember that, with the exception of sex-linked traits, two sets of DNA appear in each individual, so Mendel's laws apply.
• We must assume that the gene is randomly distributed in the population. Otherwise, there would be evidence of homosexuality in certain bloodlines, as hemophilia and sickle-cell anemia have evidence of. But homosexuality seems to appear randomly within various bloodlines, so the gene must be randomly distributed too. (Note that some "bloodlines" have a higher prevalence due to a higher prevalence of a belief in liberalism in the family, so family histories are suspect).
• We can't assume that the prevalence of the gene is equal to the prevalence of homosexuality, or the study reduces to a case of begging the question (using a statement to prove itself). That is, of course, an invalid argument method.
• We can't assume 0 for the prevalence of the DNA in heterosexuals, or there is no source to supply the inherited gene.
• We can't assume 0 for the prevalence of the trait expression in heterosexuals, or again it is begging the question.
• We can't assume 50 percent for the prevalence of the gene in the population, because genes don't work that way. Remember that there are TWO versions of each chromosome in each person (excepting the X and Y chromosomes found in a male). If they used 50 percent, they tried to use social science norms for a genetics problem.
• Possible null hypothesis values for the prevalence of randomly distributed uncorrelated genes are normally 25 percent and 75 percent, depending on the type of gene and the method of assay used. Randomly distributed sex chromosome genes have a prevalence of 50 percent in a male, and either 0 percent, 25 percent, or 75 percent in a female (depending on the location and type of gene). A randomly distributed gene location with more than two possible traits will have different percentages.
• If a gene is NOT randomly distributed in the population for some reason, then the only possible valid method of assay is to compare the population with the gene to the population without the gene, looking for percentages with the trait within each population. The null hypothesis in this case is that both populations have the same prevalence of the trait. Such a comparison was not done in either study.
• If the activity of the gene is tested for, then the laws of sexual genetics apply. A randomly distributed dominant gene has a prevalence of 75 percent, and a randomly distributed recessive gene has a prevalence of 25 percent. A sex-linked trait will have a prevalence of 50 percent in a male subject, and will be either dominant, recessive, or totally absent in a female. This generally fits the reported methods in the first study.
• If a genetic probe is used, the laws of sexual genetics apply as for a dominant gene, whether the gene itself is dominant or recessive. This fits the method of the second study.
• One last item is particularly interesting. Proving that Homosexuality is hereditary also disproves Evolution and Natural Selection. This, in turn, will cause different factions of liberals to oppose each other.

## The laws of sexual genetics as applied to expression of a gene:

Percentages are portions of each group having the gene.

Experimental group
(has trait) %
Control group
(no trait) %
Correlation
trait to gene
Type of relationship
1001000Ubiquitous uncorrelated gene
10069.5 Randomly distributed dominant gene with another random recessive gene
10060.5 Randomly distributed dominant gene with another random sex-linked gene
10043.5 Randomly distributed dominant gene with another random dominant gene
10020.5 Randomly distributed recessive gene with another random recessive gene
10014.5 Randomly distributed recessive gene with another random sex-linked gene
10008.5 Randomly distributed recessive gene with another random dominant gene
10001 Correlated gene
75750 Dominant uncorrelated randomly distributed gene
75750 DNA probed uncorrelated randomly distributed gene
50500 Sex-linked uncorrelated randomly distributed gene
50500 Environmental factor
25250 Recessive uncorrelated randomly distributed gene
0100-1 Correlated preventative gene
000 Gene not in population
xx0 Environmental factor
x100-x0 Nonrandom distribution of uncorrelated gene
x100-x0 Nonrandom sample collection

Now turn it around the way it is supposed to be, and see what happens when the independent variable is the gene, and the dependent variable is the trait:

Percentages are portions of each group having the trait.

Experimental
group (has gene) %
Control
group (no gene) %
Correlation
gene to trait
Type of relationship
1001000Ubiquitous uncorrelated trait
10001Correlated trait
75750Uncorrelated trait
75750Wrong gene (dominant)
750.5Gene requires another dominant gene to work
50500Uncorrelated trait
50500Random environmental factor
500.5Gene requires another sex-linked gene to work
250.5Gene requires another recessive gene to work
25250Uncorrelated trait
25250Wrong gene (recessive)
0100-1Negatively correlated trait
000Trait not in population
xx0Environmental factor
x100-x0Nonrandom sample collection or gene partially correlated

A randomly distributed dominant gene should appear in 75 percent of the population.

A randomly distributed recessive gene should appear in 25 percent of the population.

A randomly distributed gene should appear in 75 percent of a DNA probe assay of the population.

Any other numbers observed indicate that either the gene is NOT randomly distributed in the population, or that the sampling method is flawed.

Note that a control group is needed to be able to tell several of the cases apart.

Note that, without a control group, numbers not near 100% tend to disprove any causal claim. There are cases with the effect, but not the cause.

Note that, with a control group, numbers not near 100% disprove any causal claim. There are cases with the effect, but not the cause.

## STUDY 1

Now let's see which norms make sense from the conclusion the first study group obtained:

• If they used the actual prevalence of this DNA in the population, they would have provided that figure.
• If they had assumed the gene was randomly distributed, their conclusion would have been that the gene does not cause homosexuality.
• They might have assumed that the prevalence of the gene equals the prevalence of homosexuality. If so, the study is invalid (begging the question).
• They might have assumed 0 for the prevalence of the gene in heterosexuals.
• They might have assumed 0 for the prevalence of the trait expression in heterosexuals.
• If they had assumed a randomly distributed dominant gene, their conclusion would have been that the gene does not cause homosexuality.
• They might have assumed a randomly distributed recessive gene.

Now let's ask: Is the 26 percent a significant difference?

1. If the norm assumed was zero (the trait is not expressed in heterosexuals), then 26 percent is a value that tends to disprove the assertion, because well over half of the subjects expressed the trait without the gene.
2. If the norm assumed was the 25 percent recessive expression, then the value obtained was the possible value from a group of 15 that was closest to 25 percent. That means the gene is most likely a recessive gene that is randomly distributed among the homosexual population, and thus has no correlation to homosexuality.
3. If they did their t test on the difference between .26666... and .25, they were measuring the effects of the small sample size they used, not an actual hereditary effect. No value closer to .25 could possibly be obtained from their small sample.

In fact, when using any of these scenarios, the result disproves the assertion that the gene causes homosexuality, because the value obtained is either closest to random chance, or shows that the gene is more correlated with heterosexuality.

Note the fact that they tried to convert categorical data into the numeric values used. They used Student's t distribution, which is used to compare two numeric values, or to compare a numeric value to a norm. But they had neither two experimental values, nor did they disclose a norm to compare their one experimental count to. This test is designed to be used when the data are multiple numeric values.

But the data at hand were categorical yes-no values. The correct method is:

• Create a crosstable.
• Calculate the Chi-squared value.
• Perform a Chi-Squared test to determine whether or not there is an association.
• Cramer's coefficient can be used instead of the Chi-squared test, but it is not as accurate.

But they didn't collect the data for that.

## STUDY 2

Now let's see which norms make sense from the conclusion the second study group obtained:

• If they used the actual prevalence of this DNA in the population, they would have provided that figure.
• If they had assumed the gene probed was randomly distributed, their conclusion would have been that there was not enough evidence that the gene causes homosexuality.
• They might have assumed that the prevalence of the gene equals the prevalence of homosexuality. If so, the study is invalid (begging the question).
• They might have assumed 0 for the prevalence of the gene in heterosexuals. If so, the conclusion was invalid because they did not test for this case.
• They might have assumed 0 for the prevalence of the trait expression in heterosexuals. The conclusion is invalid for the same reason.
• If they had assumed a randomly distributed sex-linked gene, their conclusion would have been that the gene causes homosexuality. But they offered no control evidence that the gene was not just as prevalent in heterosexuals. And the gene was not on a sex chromosome.
• A randomly distributed recessive gene makes no sense with a DNA probe.

Now let's ask: Is the 82.5 percent a significant difference?

1. If the norm assumed was zero (the trait is not expressed in heterosexuals), then 82.5 percent is a value that tends to prove the assertion, but ONLY if they had collected a control group that produced a value near zero. Without the control group, the gene could be just an uncorrelated gene.
2. If the norm assumed was the 75 percent expected probed expression, then the value obtained was the possible value from a group of 40 was close enough to 75 percent to be observational error. The observed value turns out to be a difference of 3 samples, well below one standard deviation (.144) away from the expected value (.75) of a randomly correlated gene. Thus, the result is not significantly different from the norm. This is attributable to the way the sample was collected and the low number of subjects. That means the gene is most likely a gene that is randomly distributed among the homosexual population, and thus has no correlation (or a very weak correlation) to homosexuality. A control group would tell whether the gene was acting with another gene, or was uncorrelated. But they did not collect one.

In fact, using any scenario here with the second study, the result can not prove the assertion that the gene causes homosexuality without a control group to compare it to.

Note the fact that they tried to convert categorical data into the numeric values used. They used a correlation test, which is used to compare two sets of numeric values. But they did not have sets of experimental values. This test is designed to be used when the data are multiple numeric values.

But the data at hand were categorical yes-no values. The correct method is:

• Create a crosstable.
• Calculate the Chi-squared value.
• Perform a Chi-Squared test to determine whether or not there is an association.
• Cramer's coefficient can be used instead of the Chi-squared test, but it is not as accurate.

But they didn't collect the data for that.

Notice also that you can NOT select on the dependent variable (sexual preference) and expect to see the independent variable (the gene) vary in the exact manner of a causal effect. Such an experiment is designed backwards, and cannot possibly be used to prove causality. The independent variable must be actively varied and the dependent variable observed, in order to show any cause-and-effect relationship. But genes cannot be actively varied.

Notice that you can NOT use statistics intended for numeric data (each sample produces a number) when you have categorical data (yes-no or other non-numeric selectors). Statistics intended for categorical data must be used.

## MISTAKES

Horrendous mistakes were made in these studies.

• EVERY such study published so far has selected subjects for the dependent variable (homosexuality). This means that EVERY study so far is based on a logically invalid sample collection method. The samples taken should be random samples of the general population, not sets of carefully selected subjects.
• Every study published so far has no control group. This means that there is nothing to test for a correlation or an association to.
• Many of the studies, including the two examined in detail, used the methods of social science, rather than the methods of genetics, to establish the numerical values of the null hypotheses. Thus, they are trying to correlate their observations to the wrong values. Genetics has special rules that must always be followed in scientific testing. Heredity does not follow the rules of behavioral science.
• A big soduk (kudos spelled backwards) to any "scientist" who is ignorant enough to try to apply the rules of social and behavioral sciences to any genetic study.
• Because of the small sample sizes, most of the "observed variations" were caused by the limitations of obtainable values due to the discrete nature of the observations. It would take a huge variation in the variables to make a statistically significant result from such small sample sizes.
• All of the studies, including these two, used statistics intended for numeric data, even though they had categorical data. Thus, they obtained totally useless values from those statistics.
• A big soduk to any "scientist" who is ignorant enough to use the wrong statistic in any study.
• Of those who have tried to duplicate the studies, not one has succeeded in getting the same results. The wide variations are due to the small sample sizes used.
• None of them showed their data or their work. They published only the results.
• It may be significant that both websites of the research projects examined in detail were taken down (or possibly moved to different URLs) shortly after publication of this page (which originally contained links to them).
• The fact that there are cases with the effect, but not the cause, disproves their claims of causality.
• If Homosexuality is proved to be hereditary, it would disprove Natural Selection, and thus, it would disprove Evolution.

## Excuses given for using the wrong procedures in these studies:

These are the reasons often cited for the horrendous mistakes in these studies:

• The small sample sizes were used due to the expense of the DNA tests.
• The random samples needed to use the association method can be very large (~400 samples). The expense would be prohibitive.
• We followed the procedure for a causal link in our political science books.
• What's a crosstable?
• Chi squared??? What's that?
• What's Cramer's coefficient?
• But I want a correlation, not an "association". The media don't understand an association.
• We tried that. It didn't give the numbers we wanted.
• We used a 50% null hypothesis because it is supposed to be sex-linked.

Note these facts about the above excuses:

• Most of these studies were done by political scientists, who used the political science methods they learned in college.
• Very few political science courses in statistics cover how to deal with categorical data. The materials usually cover mostly environmental and social issues that use numeric values, such as pollution, population, and poverty.
• Many statistics books do not have the crosstable, Chi-squared, Cramer's coefficient, or a test for association.
• Many statistics books do not tell when to use each statistic, just how to calculate it.
• A sex-linked gene must be on the X or the Y chromosome. It can't be on any other chromosomes.
• Those funding the studies looked at the costs of tests for a minimum number of subjects when using numeric data. They would not provide any more money than that amount.

## Final conclusions:

Most of these studies were obviously designed to reach a predetermined conclusion, regardless of the actual facts. The "scientists" involved set out to prove their political beliefs, rather than to find out the truth. Otherwise, they would have used the proper scientific methods and collected a truly random sample.

As they stand, these studies have absolutely no scientific value. Instead, the people who did these studies probably fit at least one of these cases:

1. They were horribly ignorant of the laws of genetics, proper scientific procedure, proper experiment design, and/or the proper use of statistics.
2. They were trying to save money on the studies by cutting corners.
3. They were trying to prove the result they wanted, instead of trying to find out the truth.
4. They knew their results were bogus, but intended to use them to fool politicians and the press.

Apparently they had political science books on how to prove social or environmental issues, and tried to use those without any expertise.

"Studies" of this kind are meant to promote a political dogma, not to scientifically prove anything.

Numbers not near 100% tend to disprove any causal claim. There are many individual cases in these studies having the effect, but not the proffered cause. They disproved their own claims with these cases.

Beware of ANY study done by a group that would benefit from one outcome of the study. Science must be done by disinterested scientists (scientists who do not want a particular outcome to be true.)

## Epilog: The Civil Rights Entanglement of Religion and Homosexuality

Those who perpetrated these instances of bad science have as their goal a law prohibiting discrimination against homosexuality. But such a law will never exist for very long, because any such law is discrimination against religion. Discrimination against religious belief is unconstitutional in the United States of America.

Consider the following facts:

1. There is a civil right of religious belief.
2. It is unconstitutional for government to pass any law having to do with any religion.
3. It is unconstitutional for government to prohibit the free exercise of religion.
4. Most religions prohibit homosexuality as a sinful act.

For example, according to the Christian Bible, homosexual desires are caused by worshiping created things, instead of the Creator (Romans 1:25−27). This belief is protected by the Constitution. (If we assume that this is true, one might wonder if worship of the environment has caused the sudden increase in the number of homosexuals.)

Also in this example, the Christian Bible tells believers to not associate with willful sinners (I Corinthians 5:11). This belief is also protected by the Constitution.

5. Most religions prohibit changing the religious texts.
6. Many homosexuals try to argue that there is nothing to religion. But all believers know that those arguments are hollow lies.
7. Discrimination against homosexuality is against the doctrines of Political Correctness. But Political Correctness is itself a religion. Democrats plagiarized it from the tenets of the Baha'i Faith. Because Political Correctness is itself a religion, government can not use the force of law to enforce it.
8. Any attempt to make homosexuality a civil right takes away the civil rights of other people with religious beliefs.
9. Any law prohibiting discrimination against homosexuality is unconstitutional, because such a law discriminates against at least one religion (the example above), not to mention every other religion that prohibits homosexuality.
10. Any Constitutional amendment attempting to prohibit discrimination against homosexuality makes the Constitution inconsistent with itself, because it would prohibit part of itself.
11. Demands that religion be abolished are also unconstitutional.
12. No person whose religion prohibits homosexuality can be forced to pay for or provide any of the following:
• Any medical procedure or cost related to any illness caused by a sexually immoral act.
• Any cost related to any marriage license for a sexually immoral marriage
• Anything for a ceremony for a sexually immoral marriage.

Government is forcing the person to pay for these if payment is made through taxation or through required insurance premiums.

There is only one solution to this dilemma: Since religions can discriminate against each other, the only way for homosexuality to be protected is for it to become a religious belief. The Political Correctness religion would do for this purpose.