Oct 10 - Sample size

Question 1

Methods of determining Sample Size

Confidence Interval Approach

Answer 1

1) Variability
2) Confidence Interval - Probability that the true population is included in the range
(z - lvl of confidence - 1.96 or 2.58
3) Sampling Distribution
4) Standard error of a µ or %

Question 2

Standard error of a µ

Answer 2

An indication of difference between a sample value and the true population value
SEµ = sigma sample ÷ sqrt {n}

Question 3

Standard error of a %

use this for sample size calculation

Answer 3

Sp= SQRT {(p*∂)/n}

With ∂ = 1-p

Question 4

Computing Sample Size using the Confidence Interval Approach

Answer 4

A) We need:
Variability (%)
Conf lvl z (e.g. 1.96)
Acceptance error (+/- 3%)
n = {z^2 *p*∂}/ AE  

B) Estimating Sample size determinant
a) Variability: Percentage
b) Desired Accuracy (acceptable error) +/-3%
c) 95% or 99% confidence lvl is standard
95% industry norm

Question 5

Sample size, using a percentage

Answer 5

n = {z^2 p∂}/ AE

AE ~ 0.03
ALWAYS ALWAYS ALWAYS 0.03

Question 6

B) Estimating Sample size determinant

Answer 6

a) Variability: Percentage
b) Desired Accuracy (acceptable error) +/-3%
c) 95% or 99% confidence lvl is standard
95% industry norm

Question 7

Special sample size determination situations

Answer 7

• The formula is not good for nonprobability sampling

- Still good benchmark for the sample size justification

Question 8

Non-Sampling Errors

Answer 8

Include errors from:
Nonresponse, 
Data gathering, 
Data handling, 
Data analysis 
and Interpretation

Question 9

Non-Sampling Errors - Field Workers

Answer 9

Intentional: Cheating - Leading respondents

Unintentional: Misunderstanding - Fatigue - Interviewer characteristics

Question 10

Non-Sampling Errors - Respondents

Answer 10

Intentional: Falsehoods Refusals Nonresponse

Unintentional: Misunderstanding - Attention loss - Distractions - Guessing - Fatigue

Question 11

Control Measures Fields Workers

Answer 11

Intentional:
Tight supervision and monitoring
Validation of respondents’ response

Unintentional:
Orientation and training sessions
Roleplaying
Breaks

Question 12

Control Measures Respondents

Answer 12

Intentional:
Anonymity, confidentiality, Incentive, validation checks, “third-person” technique

Unintentional: 
Well-drafted questionnaire
Instructions/examples;
Reversals of scale endpoints;
Use of prompters