All of the experiments attempting to show or prove a genetic cause for homosexuality have so far been shown to be bad science. For some strange reason, the experimenters keep trying to use the same wrong methods, including:
They are continuous or interval data.
Examples:
They are discrete or nonnumeric data.
Examples:
They are positive integer (wholenumber) data.
Examples:
The purpose of science is to find out the truth, not to prove a preconceived belief.
A good study using Numerical values needs at least 30 randomly picked subjects in each category. If the data under test are Numerical values, 15 in the experimental group and 15 in the control group would do.
But in these experiments, we have only the Counts of subjects in each subset of categories. A crosstable must be used when the data are Categorical values. With Categorical values, randomly selected subjects must be collected without culling until at least 40 subjects are found for each possible value of the independent variable. In addition, at least 5 subjects should be found in each bin in the crosstable (unless none at all are found). This can result in hundreds of subjects to test. None may be removed from the sample pool without tainting the results.
One study used Student's t distribution, which is used to compare two sets of experimental Numeric values, or to compare a set of Numeric values to a norm. But they didn't have two sets of experimental values, and they did not reveal any norm.
Another study used a correlation test, but WHAT did they correlate the data to? Again, they didn't have either two experimental Numeric values or any norm to compare their one set of experimental values to.
Actually, neither of these tests is the correct test, because the data are Categorical (in this case, yesorno data). Categorical data are unordered, so there can be no Numerical value. Either the gene is there, or the gene is absent. Either the trait is there, or the trait is absent. The ChiSquared test is the only test that works with the Categorical data for both the independent and dependent variables placed in a crosstable. It compares the Counts of subjects in each category set to the expected Counts.
Since Categorical data are used, a correlation is impossible. Correlations require Numerical data. An association is tested for when both independent and dependent variables are Categorical.
If they actually understood statistics, they would have known that a causal link cannot be proved by statistics alone, and that a correlation is impossible with Categorical data.
The trouble with this is that it reduces their sample size to one sample for the remainder of any calculations they do (they wrongly kept the old sample size).
In one study, they could not correctly do long division. 4/15 is .2666  which rounds to 27 percent, not 26 percent.
In another study, they somehow got 64 percent, but no method of analysis of the Counts collected produces that value. They did not disclose the method of analysis.
One study did not reveal whether they tested for the presence of DNA or a protein produced by the gene.
Other studies said they tested for the DNA.
Others said they tested for a protein produced by the gene (more accurate).
† 
These were probably done because the government funding the researchers got was the amount provided to pay for the bare minimum number of samples needed for researching when using Numerical data. Many more samples are needed when using Categorical data. The number of tests varies, because totally random sampling must continue until each bin in the crosstable has the minimum number of samples. 
Here is the correct way to do the experiment. It must not be altered. Any change in this procedure will introduce bad science into the procedure:
One possible alternative method is to have one or more scientist with each kind of bias. They would work as checks and balances against any one bias affecting the outcome of the science.
(This can actually be tested for, because it generates valid categorical data.)
This means the subjects must be randomly selected with a method obeying the following rules:
In addition, if, after the cases are placed in the crosstable bins, any bin contains a Count between 1 ands 4, the random sampling procedure shall be continued to gain more subjects.
The subject shall NOT be shown the code number in parentheses after the items.
Crosstable  Group 1  Group 2  Group 3  Group 4  Group 5  Total  

Gene #1  Gene #2  Gene #3  Worship  Psychol  
Yes  No  Yes  No  Yes  No  Yes  No  Yes  No  
Heterosexual  0  0  0  0  0  0  0  0  0  0  0 
Homosexual  0  0  0  0  0  0  0  0  0  0  0 
Total  0  0  0  0  0  0  0  0  0  0  0 
% Homosexual  0  0  0  0  0  0  0  0  0  0  0 
Note that this is actually 5 crosstables put together, one group for each of the independent variables (top row). They are put together to prevent calculating the same value several times.
For each subject, choose one cell in the 2×2 area for each category and add one to the tally in that box:
If so, throw out all data from that subject and remove the subject from the pool. Collect another subject to replace the disqualified subject.
A Yes answer indicates that the subject is lying on the questionnaire to ruin the experiment. None of the items with code 4 actually exists.
If so, make Case A true, otherwise make Case A false.
These are cases where the person worships a created thing.
If so, make Case B true, otherwise make Case B false.
These are genuine religious beliefs.
If so, change Case B to true.
This includes the marked survey item on religious faith.
If so, make Case C true, otherwise make Case C false.
These are addictions.
If so, change Case C to true.
This includes the marked survey item on psychological causes.
If so, make Case W true, otherwise make Case W false.
This is a test for worshiping created things.
If so, make Case P true, otherwise make Case P false.
These are psychological indicators.
If so, change Case P to true.
This is a psychological indicator.
Select the row on whether or not the person indicates he is homosexual.
Select the column on whether or not the gene is present.
Add 1 to the Count in the one cell where the selected row and column meet.
Select the row on whether or not the person indicates he is homosexual.
If Case W is true, select the Worship Yes column. Otherwise, select the Worship No column
Add 1 to the Count in the one cell where the selected row and column meet.
Select the row on whether or not the person indicates he is homosexual.
If Case P is true, select the Psychol Yes column. Otherwise, select the Psychol No column
Add 1 to the Count in the one cell where the selected row and column meet.
Find the column sums, row sums, and the grand total.
Calculate the % Homosexual value by dividing the Homosexual Yes value in the column by the column sum.
The sample table with the data filled in now looks like this:
Crosstable  Group 1  Group 2  Group 3  Group 4  Group 5  Total  

Gene #1  Gene #2  Gene #3  Worship  Psychol  
Yes  No  Yes  No  Yes  No  Yes  No  Yes  No  
Heterosexual  270  90  180  180  360  0  6  354  69  291  360 
Homosexual  30  10  20  20  0  40  37  3  11  29  40 
Total  300  100  200  200  360  40  43  357  80  320  400 
% Homosexual  10  10  10  10  0  100  86  1  14  9  10 
These data were constructed solely to demonstrate five kinds of possible results, with one group showing each kind. They are not intended to show any desired result.
If any of the bins contain a nonzero value smaller than 5, then more samples must be taken.
In this case, since the value in one bin is 3, double the sample size. Collect 400 more samples. But do not increase the sample size if all deficient bins contain 0.
The new sample collection must have the same randomness as the collection taken above.
NEVER try to collect more for just that one bin. Doing this distorts the data, ruining the experiment.
The sample table with the new data added now looks like this:
These data were constructed solely to demonstrate five kinds of possible results, with one group showing each kind. They are not intended to show any desired result.
The color patches refer to calculations used below.
Crosstable  Group 1  Group 2  Group 3  Group 4  Group 5  Total  

Gene #1  Gene #2  Gene #3  Worship  Psychol  
Yes  No  Yes  No  Yes  No  Yes  No  Yes  No  
Heterosexual  540  180  360  360  720  0  10  710  140  580  720 
Homosexual  60  20  40  40  0  80  74  6  22  58  80 
Total  600  200  400  400  720  80  84  716  162  638  800 
% Homosexual  10  10  10  10  0  100  88  1  14  9  10 
Calculate the Chisquared statistic. The procedure is:
Here are the calculations. "Worship Created Things" was chosen to show the math with nonzero values:
From Crosstable 
Worship  Total  

Yes  No  
Heterosexual  720 * 84 / 800 = 75.6  720 * 716 / 800 = 644.4  720 
Homosexual  80 * 84 / 800 = 8.4  80 * 716 / 800 = 71.6  80 
Total  84  716  800 
Notice that the column and row sums still apply.
Here are the Expected Frequencies for the sample data above.
Expected Frequencies Table 
Group 1  Group 2  Group 3  Group 4  Group 5  Total  

Gene #1  Gene #2  Gene #3  Worship  Psychol  
Yes  No  Yes  No  Yes  No  Yes  No  Yes  No  
Heterosexual  540  180  360  360  648  72  75.6  644.4  145.8  574.2  720 
Homosexual  60  20  40  40  72  8  8.4  71.6  16.2  63.8  80 
Total  600  200  400  400  720  80  84  716  162  638  800 
Here are the calculations. "Worship Created Things" was chosen to show the math with nonzero values:
Partial ChiSquared 
Worship  

Yes  No  
Heterosexual  ( 10 − 75.6) ^ 2 / 75.6 = 56.92  ( 710 − 644.4) ^ 2 / 644.4 = 6.68 
Homosexual  ( 74 − 8.4) ^ 2 / 8.4 = 512.30  ( 6 − 71.6) ^ 2 / 71.6 = 60.10 
The colors show where the values are obtained from in the Crosstable and the Expected Frequencies Table above and where the results are put in the Partial ChiSquared Table below.
Here are the Partial ChiSquared cell values for the sample data above.
Partial ChiSquared 
Group 1  Group 2  Group 3  Group 4  Group 5  

Gene #1  Gene #2  Gene #3  Worship  Psychol  
Yes  No  Yes  No  Yes  No  Yes  No  Yes  No  
Heterosexual  0.00  0.00  0.00  0.00  8.00  72.00  56.92  6.68  0.23  0.06 
Homosexual  0.00  0.00  0.00  0.00  72.00  648.00  512.30  60.10  2.08  0.53 
Here is the calculation. "Worship Created Things" was chosen to show the math with nonzero values:
56.92 + 6.68 + 512.30 + 60.10 = 636.00
Here are the final ChiSquared Statistics for the sample data above.
Final ChiSquared 
Group 1  Group 2  Group 3  Group 4  Group 5 

Gene #1  Gene #2  Gene #3  Worship  Psychol  
ChiSquared Statistic  0.00  0.00  800.00  636.00  2.90 
Cramer's Coefficient converts the ChiSquared value into a proportion or a probability.
Calculate Cramer's Coefficient from the final ChiSquared values.
Here is the calculation. "Worship Created Things" was chosen to show the math with nonzero values:
√(636.00 / 800 / (21)) = 0.89
Here are the Cramer Coefficients for the sample data above.
Cramer's Coefficient  Group 1  Group 2  Group 3  Group 4  Group 5 

Gene #1  Gene #2  Gene #3  Worship  Psychol  
Cramer's Coefficient  0.00  0.00  1.00  0.89  0.06 
The procedure is:
The value retrieved for .05 significance and 1 degree of freedom is 3.84.
Finish the ChiSquared table with the significance level, and whether or not the ChiSquared value exceeds it.
ChiSquared  Group 1  Group 2  Group 3  Group 4  Group 5 

Gene #1  Gene #2  Gene #3  Worship  Psychol  
ChiSquared Value  0.00  0.00  800.00  636.00  2.90 
ChiSquared distribution value  3.84  3.84  3.84  3.84  3.84 
Is group significant?  NO  NO  YES  YES  NO 
Alternately, use Cramer's Coefficient:
Cramer's Coefficient Range  Meaning 

1.0 .. 0.8  Strong Association 
0.8 .. 0.5  Fair Association 
0.5 .. 0.25  Weak Association 
0.25 .. 0.0  No Association 
Cramer's Coefficient  Group 1  Group 2  Group 3  Group 4  Group 5 

Gene #1  Gene #2  Gene #3  Worship  Psychol  
Association Strength  0.00  0.00  1.00  0.89  0.06 
Case Association Level  None  None  Strong  Strong  None 
In the trial exercise above, 5 different cases of recognizing association are shown:
In this case, ChiSquared and Cramer's Coefficient show no association.
Note that the gene is active in 75% of the cases, both homosexual and heterosexual.
Suspect another cause, because the gene is not associated.
Variations:
Unassociated Recessive Gene:
Note that the gene is active in 25% of the cases, both homosexual and heterosexual.
Types of tests:
If the test is for the product of the gene, the numbers above are correct.
If the test is for the gene code, the test numbers will be 75%, whether dominant or
recessive. This kind of testing should be avoided if the gene is recessive,
In this case, ChiSquared and Cramer's Coefficient show no association.
If both of the dependent values show the same percentage of the gene, but the percentages are not 75%  25% or 25%  75%, suspect an unassociated environmental value, a nonrandomly distributed gene, or a biased sample.
If both of the dependent values show the same percentage of 50% of the gene, an unassociated sexchromosome gene might be the cause of the numbers.
In this case, ChiSquared and Cramer's Coefficient show a strong association.
Since the presence of the gene is perfectly associated with heterosexuality, and the absence of the gene is perfectly associated with homosexuality, the gene probably prevents homosexuality. But without a causal analysis, this is only a negative association. Pursue more on this factor.
Variations:
A positive association would have the nonzero Counts in the Homosexual Yes and Heterosexual No cells.
In this case, ChiSquared and Cramer's Coefficient show a strong association.
Since the calculated ChiSquared greatly exceeds the distribution table value, the association is very strong, but a few cases do not follow the association. Pursue more on this factor.
In this case, ChiSquared and Cramer's Coefficient show no association.
Since the calculated ChiSquared value is less than the distribution table value, the association is weak, if not nonexistent. This factor is not interesting enough to pursue.
The above data were constructed solely to demonstrate these five kinds of possible results, with one group showing each kind. They are not intended to represent any truth in this controversy.
A few pitfalls to avoid in studies of this type:
Although bias cannot be totally eliminated, efforts to prevent bias or compensate for it can reduce the bias.
The page author would be interested in reading any study conducted the correct way. Go to his home page to find contact information.
In this case, one table is made for each possible cause. It is well suited to being used with Excel.
The sample table below is the same data as that contained in the sample table with new data above
These data were constructed solely to demonstrate possible results. They are not intended to show any desired result.
The color patches refer to the sources of the values used in the calculations.
Group 4 Worship 
Crosstable  Expected Values 
Partial Chi Squared 
Cramer's Coeff 


Yes  No  Total  Yes  No  Total  Yes  No  Stat  Stat  
Heterosexual  10 
710 
720 
720
*
84
/
800
= 75.6 
720
*
716
/
800
= 644.4 
720 

Homosexual  74 
6 
80 
80
*
84
/
800
= 8.4 
80
*
716
/
800
= 71.6 
80 

Total  84  716  800  84  716  800 
Group 4 Worship 
Crosstable  Expected Values 
Partial Chi Squared 
Cramer's Coeff 


Yes  No  Total  Yes  No  Total  Yes  No  Stat  Stat  
Heterosexual  10 
710 
720 
75.6 
644.4 
720 
( 10 −
75.6) ^ 2 /
75.6
= 56.92 
( 710 −
644.4) ^ 2 /
644.4
= 6.68 

Homosexual  74 
6 
80 
8.4 
71.6 
80 
( 74 −
8.4) ^ 2 /
8.4
= 512.30 
( 6 −
71.6) ^ 2 /
71.6
= 60.10 

Total  84  716  800  84  716  800 
Group 4 Worship 
Crosstable  Expected Values 
Partial Chi Squared 
Cramer's Coeff 


Yes  No  Total  Yes  No  Total  Yes  No  Stat  Stat  
Heterosexual  10 
710 
720 
75.6 
644.4 
720 
56.92 
6.68 
56.92
+ 6.68
+ 512.30
+ 60.10
= 636.00 

Homosexual  74 
6 
80 
8.4 
71.6 
80 
512.30 
60.10 

Total  84  716  800  84  716  800 
Group 4 Worship 
Crosstable  Expected Values 
Partial Chi Squared 
Cramer's Coeff 


Yes  No  Total  Yes  No  Total  Yes  No  Stat  Stat  
Heterosexual  10 
710 
720 
75.6 
644.4 
720 
56.92 
6.68 
636.00 
√( 636.00 / 800 / (21)) = 0.89 
Homosexual  74 
6 
80 
8.4 
71.6 
80 
512.30 
60.10 

Total  84  716  800  84  716  800 
Group 4 Worship 
Crosstable  Expected Values 
Partial Chi Squared 
Cramer's Coeff 


Yes  No  Total  Yes  No  Total  Yes  No  Stat  Stat  
Heterosexual  10 
710 
720 
75.6 
644.4 
720 
56.92 
6.68 
636.00 > 3.84 
0.89 
Homosexual  74 
6 
80 
8.4 
71.6 
80 
512.30 
60.10 

Total  84  716  800  84  716  800  Assoc  Strong 