Supplement 5

Many studies are said to have proven the relationship between blood types and personality.  But three problems shown below are laying ahead.

1. There isn't reliability of data, because random sampling method is not carried out.
2. There is little reappearance of data.  Trends differ by questionnaires.
3. If trends appear, they are different from blood-typical personality.

If one can solve these three problems, we can say "the relationship between blood type and personality is proved completely," as Alexander the great cut the Gordian Knot.

I think prerequisites like the following are necessary to get stable results. 

1. Homogeneous subjects (social position, age, gender, region etc.)
2. The number of subjects is more than several hundred (more than one thousand and subjects of each blood types are the same, if possible)
3. Choose blood-typical personality question items (do not use psychology personality tests)
4. Results do not directly correspond to blood-typical descriptions ("words" do not represent real "personality") -- also affected by culture, nationality etc.
5. Consider statistical errors carefully - they are much larger than usually thought.

There appears only 10 - 20% differences -- at most -- of the results of questionnaires by blood types.  Needless to say, differences depend on contents of questions.  Moreover, there are fairly many factors that affect trends: social position, age, sex, region etc.  Even the same blood-type person answered differently by these factors.  So, when "random sampling" is carried out, these factors sometimes overwhelm blood types.  In many cases, differences do not come out.  Never use "random sampling" which is the popular method of psychology!

So, clear differences have appears many times, when we look at data of university students.   Because there is few differences -- such as social position, age, sex, region -- very homogeneous data. 

As I have mentioned, there appears only 10 - 20% differences, at most. Significant results of χ2-test will not appear if the number of subjects is less than several hundred.  More than 1,000 subjects are recommended, if possible.

This is natural.  Because differences are difficult to come out in usual personality test.  Question items of personality tests themselves are hard to measure the differences by blood types.  I don't know the true reason yet ... 

As I have mentioned, there are fairly many factors that affect trends: social position, age, sex, region etc.  Even the same blood-type persons answered differently by these factors.  "Words" do not represent direct "personality" -- also affected by culture, nationality etc.

 They are much larger than you might think.  For example, a statistical error (confident interval of yes-no question item) of 3,000 subjects is more than 7%!  Don't you think this is pretty large?