Calculating Sampling Error

1
(No Transcript)
2
Topic 13 Calculating Sampling Error And Sample
Size
3
Oversampling

• Why?
• How much?
• Depends on
• Where do you find completion rate?

4
Oversampling calculation

• If N300 and completion rate is 52 (.52)
• 300/.52576.9577
• If N600 and completion rate is 45 (.45)
• 600/.451333.3 1334

5
Sampling error

• random samples have sampling error
• Polls say 4 Example Pew Center
• Remember, were dealing with probability
• Rule of thumb
  ,
• which means sure that the number
  produced by the sample is accurate within a
  certain range called the
  .

6
Max. sampling error--calc1

• tells us
  what the biggest error is BEFORE conducting the
  study.
• SEmax 1.96? (.25/N)
• If expected N350
• SEmax 1.96? (.25/350) .052 or
• 5.2

7
Max. sampling error--Why?

• WHY?
• Another example assume N1200
• SEmax 1.96? (.25/1200) .028 or
• 2.8

8
Max. sampling error--Meaning

• Numbers are
• If you need accuracy you can tell if your sample
  will get the accuracy you want
• Tight political race requires
• Very competitive market requires
• Rough estimates require
• Initial market descriptions require

9
Question

• Put this on a piece of paper
• Day of the week, Todays date
• Your name
• Your signature
• Topic 13

10
Sampling error--calc

• tells us what the error is
  for a result AFTER conducting the study.
• SE 1.96? (p x q)/N
• p is the estimate (e.g. 54.54)
• q is (1-p)

11
Sampling error--example

• If 52 support president N400
• p .52 ( 52.52)
• q (1-.52) .48
• SE 1.96? (.52 x .48)/400
• .04896 or 4.9

12
Sampling error--Meaning

• Numbers are only . SE gives us
  accuracy of estimates.
• Not exactly 52 support the president, but we
  are sure that
• 52
  OR
• Between support the
  president
• Questions
• 1. Does a majority support the president?
• 2.

13
What does this mean?

• Random samples the population
• Results are
• Dont
• Dont
• Dont
• Dont

14
Example of meaning

• After a campaign to develop trust in your
  company, you conducted a study (N250) in which
  you found that 55 have high trust in your
  company and 44 have high trust in your
  competitor. Thats 11 difference!
• QUESTION Are you sure you are ahead of your
  competition?

15
Example of meaning (cont.)

• Competitor 44 say they have high trust in your
  competitor
• SE"6.2
• Range of accuracy
•  
• You 55 say they have high trust in
  you
• SE"6.2
• Range of accuracy

16
More examples?
17
Sample sizes and SE

• If N100, SEmax
• If N200, SEmax
• If N400, SEmax
• If N800, SEmax

18
Sample size/SE relation

• Question what is the relation between sample
  size and sampling error

19
Sample size/SE relation (cont.)
20
So what is the best size?

• All other things equal,
• Below that is
• Above that is

21
Calculating exact N

• N .96/(SE)2
• Assume youll accept "2.5 error (SE".025)
• N .96/(.025)2
•   N 1536
• Oversample, assume 43 completion rate
• Real N 1536/.43 3572.09 or 3573

22
Question

• Assume 48 have heard of your company (based on a
  survey of N375). What is the sampling error
  (just set up the calc)?

23
Next time

• Administering the study

24
(No Transcript)