Six Sigma

Discrete Probability Spaces

Discrete probability spaces:

Before  continuing  with  the  discussion  of  σ-fields  and  probability  measures  in  their  full  generality,  it  is  helpful  to  consider  the  simpler  case  where  the  sample  space  Ω is  finite  or  countable.

For  simplicity,  we  will  usually  employ the  notation P(ω) instead  of  P({ω}),  and  we  will  often  denote  P(ωi ) by  pi

The following are some examples of discrete probability spaces.  Note  that  typically  we  do  not  provide  an  explicit  expression  for  P(A) for every  A ⊂Ω.  It  suffices  to  specify  the  probability  of  elementary  outcomes,  from  which  P(A) is  readily  obtained  for  any  A.

Examples:

(a)  Consider a single toss of a coin. If we believe that heads (H) and tails (T) are equally likely, the following is an appropriate model. We set Ω = {ω1, ω2), where ω1 = H and ω2 = T, and let p1 = p2 = 1/2.  Here,  F = {Ø, {H}, {T}, {H, T}},  and  P(Ø) = 0,  P(H) = P(T) = 1/2,  P({H, T}) = 1.

(b)  Consider a single roll of a die. If we believe that all six outcomes are equally likely, the following is an appropriate model.  We set Ω = {1, 2 . . . 6} and p1 = · · · = p6 = 1/6.

σ-FIELD:

1.       When  the  sample  space  Ω is  uncountable,  the  idea  of  defining  the  probability  of  a  general  subset  of  Ω in  terms  of  the  probabilities  of  elementary  outcomes  runs  into  difficulties. 

2.       Suppose,  for  example,  that  the  experiment  consists  of  drawing  a  number  from  the  interval  [0, 1],  and  that  we  wish  to  model a situation  where  all  elementary  outcomes  are  “equally  likely.”  If  we  were to  assign  a  probability  of  zero  to  every  ω,  this  alone  would  not  be  of  much  help  in  determining  the  proba­bility  of  a  subset  such  as  [1/2, 3/4].

3.       If  we  were  to  assign  the  same  positive value  to every  ω,  we  would  obtain  P({1, 1/2, 1/3, . . .}) = ∞,  which  is  undesirable.

4.       A way  out  of  this  difficulty is  to work  directly  with  the  probabilities  of  more  general  subsets  of  Ω (not  just  subsets  consisting  of  a  single  element).