Week 3

Validity

• how well a measuring devise measures what is it supposed to.
• how well the operational definition matches the conceptual definition

Reliability 

• a measuring is reliable if it consistently gives us the same answer.
• measured through stability, internal consistency and equivalency 
• is a necessary condition for validity
• = operational definition

 

Kinds of error in Operational Definitions:

1. systematic
2. random

Systematic Error

• consistently not measuring what we think we are.
• operational definition doesn't represent conceptual definition well
• problem with validity

Random Error

•  if random error in operational definitions is controlled, the measure (exam/survey) will be more reliable.

 

 

μ   = POPULATION MEAN

x̄  or σ  = SAMPLE MEAN

N   = POPULATION

n   = SAMPLE

 

Population

• entire group or class group being studied

Sample

• subjects of population that represents its population well

Parameter

• value that describes a population

Statistic

• value that describes a sample

 

Descriptive Statistics:

[mean] statistical measure that takes one score that represents the information as a whole 

 

Inferential Statistics:

techniques that allow us to study a sample then generalize it to the population

   —can help us generalize about the populations from which samples were selected

   —can tell us how likely it is that the differences between samples reflect real differences between populations

 

 

 

Non Probability Sampling (cannot calculate sampling error, try to avoid)

•  chose who is available [doesn't necessarily represent population you are talking about]
  • unqualified volunteer sample   
  • purposive sample [survey those who relate to survey]
  • quota sampling
  • snowball sample [survey those who know people who have taken it]

 

Probability Sampling (can calculate sampling error)

•  random sampling: everyone has an equal chance of being selected to respond
  • systematic random sampling
  • stratified sampling
  • multistage sampling

 

Central Limit Theorem

the average/sum of a large bunch of measurements follows a normal bell curve even if the individual measurements do not

         applies to distribution of sample means for ANY population

         don't need large sample size before distribution is almost perfect (n=30)

         can tell how much sampling error we have based on the relationship between samples and the normal curve