Sampling and Measurement
I. Sampling
A. Samples v. Populations
Sample = group of people participating in the study - is supposed to be representative of whole population
Population = group of people to whom you want to generalize your results.
- Target Population : who do you really want to generalize your results to.
- Accessible Population : who you can actually generalize your results to.
ie: want to study improving test scores in Utah, but can only survey 100 teachers in Salt Lake County
target population is all teachers in Utah
accessible population is teachers in Salt Lake County, because teacher issues in rural areas are different than urban areas.
Target and Accessible populations depend on how well your sample really represents your TP. I your sample really represents your TP well, then AP and TP are the same. If sample does not represent TP well, then AP is who the sample really represents.
.
Two types of Sampling
.
1. Probability Sample = take a random sample from the population - each member of the population have the same chance of being selected as every other member. (Simple or Straight Random Sample)
.
2. Non-Probability Sample - or non-random sample = where members of the sample are chosenin a way that not every member has an equal chance of being chosen.
.
Probability Sampling Methods
.
1. Stratified Random Sampling - selected subsets of the population are chosen to represent the same proportions as in the general population. ie: x% male - y% female, or by race x% white, y% Hispanic, z% Asian etc..... Make sure the sample has the same percentages as the general population.
.
2. Clustered Random Sample - select existing groups of participants instead of creating subgroups. ie: making sure that you have a sample from both lower and higher socio-economic levels, from naturally forming groups (east v. west side schools), without forming the groups, without worrying about % representation.
.
3. Two Stage Random Sampling = combines stratified and clustered sampling. First you pick the naturally occurring groups to sample from, and then instead of using the entire population of the groups you take a sample from the clusters that were chosen.
.
Non-Probability Sampling Methods
.
1. Systematic Sampling = every nth individual in a population is selected for participation in the study. Polling every 4th person that leaves the voting booth. Sampling Interval = n and in the nth person.
.
2. Convenience Sampling = select a group of individual who are conveniently available to be participants in the study. ie: everyone at the Chevron. ok if you are studying people who shop at convenience stores at the given time - but not really a good sample of the entire population of Salt Lake City.
.
3. Purposive Sampling = researchers use past knowledge or own judgement to select a sample that he/she thinks is representative of the population.
.
Sampling in Qualitative Research
.
- Purposive Sampling - ie: teacher burnout - what do you mean by teachers, what do you mean by burnout and what do you mean by teacher burnout. - select those individuals that researcher thinks reperesents the desired population.
.
- Case Analysis
.....-typical - what typical teacher burtnout looks like - ie any teacher sometime in April or May
.....-extreme - the teacher who burned out so bad they quit and went to work for Blockbuster
.....-critical - critical characteristics of burnout - same lessons, notes laminated, don't care anymore, just waiting for retirement.
.
- Maximum Variation - sample represents maximum extremes of your population. ie: from Hickman to
.
- Snowball Sampling - start out with small group and keep adding on as the study continues.
Sampling and Validity
What size is appropriate?
Descriptive = 100
Correlational = 50
Experimentsl = 30
How generalizable is the sample?
external validity = the results should be generalizable beyond the conditions of the study? Is it only valid for the study sample?
1. Population generalizability - degree to which results can be extended to other populations
2. Ecological generalizable - degree to which results can be extended to other settings or conditions
What is Measurement
-Measurement - just the gathering of information
-Evaluation - making judgements from the collected data
-Where doe assessment fit in?
What kind of scale is the measurement based on?
-Nominal - categorical variables - gender, eye color, type of car - only one that is qualitative not quantative
-Ordinal - ranked - 1st, second, third etc - no info on how much distance between 2nd and 3rd,etc...
-Interval - no absolute zero - degrees farenheit
-Ratio - there is an absolute zero - degrees kelvin
interval and ration tend to look the same
Types of Educational Measures
- Cognitive (how much did someone learn?) v. Non-cognitive
- Commerical (standardized tests developed for commercial purposes, good, tested, norming evidience, but not tailored to the study) vs. Non-commercial (developed by researcher for their study)
- Direct (participants themselves are giving the information) v. Indirect (information about students from teachers)
Sample Cognitive Measures
-Standardized Tests
---acheivement tests - what have people already learned
---aptitude tests - potential for future learning
-Behavioral Measures
---naming time
---response time
---reading time
------wpm
------eyetracking measures - exactly where they are looking on a computer screen, how long they look there, where they go next and if they come back - images v. text
---number of fixations on Areas Of Interest
---transitions between AOI
---Duration if individual fixations or combinations of fixations on AOI
---regressions in or out of AOI
---rereading of AOI
---pupil diameter - measure of cognitive level - interested, scared, aroused, working hard on something our pupil becomes bigger.
.
Non-Cognitive Measures
---surveys and questionnaires
---observations
---interviews
.
.
How is an individual's score interpreted?
---Norm-refrenced - grading on a curve, an individuals score based on comparison to peer scores
---Criterion-referenced
.
Interpreting scores
Different ways to present scores
1. Raw scores - number of items answered correctly, number of times behavior is tallied
2. Derived scores - scored changed into a more meaningful unit
---age/grade equivalent scores -
---percentile ranking - ranking of score compared to all other individuals who took the test.
---standard scores (z-score) - how far scores are from a reference point; usually best to use in research.
mean 560 compared to
mean 140 and students score is 132
z score basically same as standard deviations. z score of 3 is 3 standard deviations from the mean.
Important Characteristics of Measure
---must be objective
---have to be useable - if they are insanely difficult to use then they are no good.
---validity - does measure actually measure what it is supposed to measure?
---reliability - do I get consistant measures over time.
No comments:
Post a Comment