I would like to preface this note with a disclaimer - I am not a statistician!!!
What I do have are two graduate stats and research courses.  I could be
dangerous.....  :-)

Rachel :   You have asked the million dollar question - which factor leads to
another.  That is what researchers attempt to determine everytime they collect
and crunch their data.  In breastfeeding vs formula research terms ,
breastmilk/feeding is the independant variable (cause) and the outcome(effect)
measure (health, inteligence, disease incidence, etc ) is the dependant variable
- SES is an extraneous variable( extraneous variables aka intervening,
confounding, demographic, organismic, or environmental variables ).  Extraneous
variables  are variables that can interfere with the action of the variables you
are studying.  An example  - I am studying the effect of epidural anaesthesia on
breastfeeding activity in the first week postpartum.  I know from a literature
review that SES may also have an effect on breastfeeding longevity.    I need to
control for that in my research design or in my data analysis so I can comment
on the effect A (epidurals) have on B (breastfeeding activity).  I dont want C
(SES) interfering with the relationship that may or may not exist between A and
B.    Have I lost anyone yet????
I can control this variable when I do my research in a few ways.  I can admit
the same number of epidural patients within a number of SES groups and MATCH
these with the same percentage in each SES women who do NOT have epidurals.
This is to say - If I have 50 women who have epidurals and 50 who do not - I
should have the same  percentage from low, mid and upper SES in each group.
Matching - according to my professors - is not a great research technique and
should be avoided.  Randomization is a very  good research technique - but not
ethical in this particular research question.  Or - I can choose a sample which
is homogenous - it does not vary (ie - all women in the sample will be from X
socio-economic status) This method has a big drawback - you can not generalize
your findings to the general population.  It only applies to X SES.  In this
case I could say - in women who are middle class - babies were less likely to
take more than 24 hours to effect breastfeeding if they had no anaesthetic when
compared to babies whose mothers received epidurals (then I would have to give a
value telling the statistical significance of these findings - and you would nod
knowingly - saying "what a great researcher Kim is...")
Another method involves using a statistical (MATH) analysis method to try and
CONTROL the effect it may have on my dependant variable.  OR you can calculate
the statistical significance of the variable (SES) between the groups ( in mine
it would be = is there a statistically significant difference in parental  SES
of  babies breastfeeding well vs not bfing well )  If the answer _ after the
simple math -  is NO - and I satisfied some stat rules - then I could go on and
comment on the relationship between A and B without worrying about whether C is
to blame.

Oh boy - I know I have LOST all you out there who have never had the pleasure of
research methods and statistical analysis.....

So What am I getting at?   When reading research articles it is important to
review the researchers methods and data analysis techniques to determine what
measures were undertaken to speak to the extraneous variables.  Did they look at
SES when looking at the incidence of juvenile onset diabetes?  Is there a
relationship between SES and juvenile onset diabetes?   DId the researchers look
at SES and its relationship to Otitis Media?  If they did find a relationship -
did they also look at smoking?  Was SES related to smoking?  If smoking was also
high then - maybe it wasnt' The SES - but the smoking - that contributed to the
Otitis Media...  See how complicated it can get????  In reality - the  data on
SES does not support a strong relationship - there is a lot of data out of the
third world -  where almost everyone studied lived in poverty - and
breastfeeding provided great rewards.  In North America - our data is more
biased the other way - many of our research subjects are not the
underpriveledged, but rather the educated and priveledged.   A researcher must
always tell us from where the sample population was obtained and how they were
obtained.   This helps us look for bias.  Always be critical in your analysis.


For people who like to say that it is all related to SES - ask them to show you
the data which supports that. - The onnus should be them to support their
position as well.

Thanks for reading this far - go have some chocolate now.    Sorry to say - I
love this stuff (research and stats) but I know I am weird.....
I will go back to my archives - because I do have info on SES - but for now I
rely of grey matter - I don't think SES has been found to be highly correlated
with many of the benefits attributed to breastfeeding.  This means - that when
all things are considered - breastfeeding benefits are still there regardless of
SES.

I apologize for the foggy discourse on research.  I leave you with one more
comment.  I am often disturbed to read about people interpreting research with
the statement that "it was proven".   In human research it is extremely
difficult to prove anything.  What we find are relationships.  Some
relationships are very strong  -  like smoking and lung cancer.  some are much
weaker - like stomach sleeping and SIDS .  In order to be more "correct"  we
should learn to say - "there is a strong relationship between breastfeeding and
subsequent reduction in maternal breast cancer".

Kim Campbell RN MN(c) IBCLC   Vancouver. ( the fanatical graduate student - give
me data or death....)