Paper: Revisiting a 90-year-old debate: the advantages of the mean deviation

Actually im busy doing an exam paper for linguistics class, but it turned out to be not so difficult, so i spent som time on Khan Academy doing probability and statistics courses. i want to master that stuff, especially the stuff i dont currently know the details about, like regression.

anyway, i stumpled into a comment asking about the way the standard deviation is calculated. why not just use the absolute value insted of squaring stuff and taking the square root after? i actually tried that once, and it gives different results! i tried it out becus the teacher’s notes said that it wud giv the same results. pretty neat discovery IMO.

anyway, the other one has a name as well:

here’s a paper that argues that we shud really return to the MD (mean deviation). i didnt understand all the math, but it sure is easier to calculate and the meaning of it easier to grasp, altho its probably too difficult to switch now that most of statistics is based on the SD. still cool tho.

Revisiting a 90-year-old debate the advantages of the mean deviation

ABSTRACT:  This  paper  discusses  the  reliance  of  numerical  analysis  on
the  concept  of  the  standard  deviation,  and  its  close  relative  the  variance.
It  suggests  that  the  original  reasons  why  the  standard  deviation  concept
has  permeated  traditional  statistics  are  no  longer  clearly  valid,  if  they
ever  were.  The  absolute  mean  deviation,  it  is  argued  here,  has many
advantages  over  the  standard  deviation.  It  is more  efficient  as an
estimate  of  a population  parameter  in  the  real-life  situation  where  the
data  contain  tiny  errors,  or  do  not  form  a completely  perfect  normal
distribution.  It  is  easier  to  use,  and more  tolerant  of  extreme  values,  in
the  majority  of  real-life  situations  where  population  parameters  are  not
required.  It  is  easier  for  new  researchers  to  learn  about  and  understand,
and  also  closely  linked  to  a number  of  arithmetic  techniques  already
used  in  the  sociology  of  education  and  elsewhere.  We  could  continue  to
use  the  standard  deviation  instead,  as we  do  presently,  because  so  much
of  the  rest  of  traditional  statistics  is  based  upon  it  (effect  sizes,  and  the
F-test,  for  example).  However,  we  should  weigh  the  convenience  of  this
solution  for  some  against  the  possibility  of  creating  a much  simpler  and
more  widespread  form  of  numeric  analysis  for  many.

Keywords:  variance,  measuring  variation,  political  arithmetic,  mean
deviation,  standard  deviation, social  construction  of  statistics

it also has a new odd use of “social construction” which annoyed me when reading it.

Leave a Reply