INFO 370 - L07 - October 20, 2004
Notes by: Aaron, Yaptinchay, Sean, Prins
Tomorrow's lab to start @ 5PM
Google Talk in EE105
Mean greater than median on a positive skew
Median greater than the mean on a negative skew
Central Limit Theorem:
> Various distributions and means were taken from samples
> 70-80 observations gives you a pretty good estimate of the mean
// Begin Slides
Generalization
> Sample generalization: form a sample to the whole population
ex. using info students to generalize UW students
> Cross-population generalization: from one population to another
ex. using UW students to generalize other university students
Evaluating Sampling Quality
> How clearly was the population defined?
> What methods were used to select the sample?
> Do the cases selected represent the population from which they were selected?
> Is cross-population generalization valid?
Factors Acting Against Quality
> The sampling frame is incomplete
- ex.: "Dewey defeats Truman" newspaper flop
- people may have been left out
- cell phone users nowadays
> Low participation rate
- self-selection bias
- make sure your questions aren't too probing, personal, turns off and stops answering question
- once subjects gets past 50% of the survey, momentum makes them finish and they are more willing to answer tougher questions
> Unwarranted cross-population generalizations
- include all socio-economical biases
Sample Size
Factors determining the selection of size:
> The degree of confidence required
> The homogeneity of the population (they all respond in a similiar fashion)
> The complexity of data analysis
Conclusion
> Selecting units in such a way that what we learn about the sample holds true for the population
> Sampling method determines the degree of generalizability
Random Samples *CHECK*
Non-random samples *X*
# END #