Randomness, Samples Surveys, and Observations Flashcards
Pick numbers for lottery
Random
Start with seed value and build with algorithm Random integer table -Appendix F -Start at top left -Read 1 digit, 2, 3 ect. -If not in range, then skip
Pseudo random
Math Prob rand is [0,1] randInt(min, max, number of answers) -Can store into list -Not equal probability
Random integer
Dice Rolling (2)
- 6 options for one dice
- 6 options to pair it with
- 36 combinations
Calculating possibilities
Math
Remainder (x,y)
Remainder
Apps
ProbSim
Simulate dice
Assign random integers
Produce random integers
Sort
Store into list
Comparing
Sample should be _ and represent whole population which can be obtained by _.
Diverse, randomizing
_ rather than choice in sampling.
Random
Key numbers in mathematical models
Population parameters
Means that the statistics reflect the parameters
Representative
What sample size in drawn from
Sampling frame
If sample is different, then data will be different
Benefits stratified
Sampling Variability
Simple Random Sample
- Has to be equivalent of drawing names out of a hat
- Every subset of size n has _ chance of being chosen
- Not every individual has an equal chance (only _) because any sampling method has an equal chance for the individual
Equal, subset
Stratified (like a layered cake)
- Choose the separation
- Layers are as thick as that group is _ in the population
- Predefine groups (_) and perform an SRS on each strata, then put them apart and back together
- Better than SRS because more representative of population
Represented, strata
Systematic
-Choose every _ term
nth
Cluster Sampling
- Population is already split up into _
- Choose a cluster(s) by SRS
- Ex. School, zip code, street
- Perform a _ of that cluster
- A census is asking or measuring everything
- Questioning the whole population
- Good when _ separated
Clusters, census, physically
Convenience Sample
- Voluntary response survey
- Usually a lot of _
Bias
Combine several sampling methods
Multilayer samples
Trail run of survey
Pilot
Only people with strong opinions respond
Voluntary response bias
Those who don’t care don’t respond
Nonresponse bias
Not proportionally representative
Under coverage
Something is done to influence people’s responses or they give incorrect answers
Response bias
Characterized by giving treatment and random assignment and see cause and effect
Experiments
Stratified: _
Blocking
Only block if there is a _ in the results based on the separation
Difference
Completely Randomized (Good)
- Treatments assigned randomly to all _
- Compare effects of treatment
Subjects
Block Design (Better)
- Splitting subjects into heterogeneous groups (blocks)
- Assign treatments randomly in each _
- Compare effects across blocks
Block
Match Pairs (Best) -Every subject receives _ treatment or treatment and placebo
Both
Factor
Broader
Level
Extra division of factor
Attributes to effect
Response
Subjects
Experimental units
Designing Experiments
- Assign random _ on call list
- Block if necessary
- Generate random integers
- Skip _
- Skip numbers not in _
- Associate back to names
- Call them
- Try to have at least 5 trials
- Should have a _
Integers, repeats, range, control
Subject doesn’t know what treatment
Blind (Good)
Subject and evaluator don’t know which treatment
Only head scientist knows
Double Blind (Better)
Levels of one factor are associated with another factor in such a way that the effects cannot be separated
Confounding
Pairing people who are similar to see results of different variables
Matching
No treatments or manipulations just study records
No cause and effect
Only correlation and association
Observational studies
_
- Past
- Find relationships
- No need to be randomized
Retrospective
_
- Future
- Find relationships
- Estimate differences
Prospective
