Use LEFT and RIGHT arrow keys to navigate between flashcards;
Use UP and DOWN arrow keys to flip the card;
H to show hint;
A reads text to speech;
20 Cards in this Set
- Front
- Back
Reliability
|
- Does my measure consistently respond the same way to the same situation?
- Consistency and repeatability |
|
Ways of checking reliability:
Test-Retest Reliability |
- Same test, two time points
- Same measurement to finest detail - Do not want to do this is if there is a chance for things to change between time one and two Ex) Step on scale, step off scale |
|
Ways of checking reliability:
Inter-Rater Reliability |
- Same test, done by two different people
- Different versions of measurement technique (same time point or diff. time points) - Want to make sure you get the same results - Important with face-to-face interviews when the person collecting the data is engaged and involved |
|
Ways of checking reliability:
Inter-Method Reliability |
- Two different tests, developed by the same methods
- Say question again to people that isn't as familiar - There should be a correlation b/t the two questions - Changing wording or phrases |
|
Ways of checking reliability:
Internal Consistency |
- Compare different items on the same test
|
|
Comparing two tests using correlations starts with...
|
- Sample 1 on x axis
- Sample 2 on y axis |
|
If increasing slope, then there is ___________________.
|
Consistency
|
|
R =
|
- How strong the correlation is
- How tight/how clustered the data is around the line of best fit |
|
P =
|
How significant the correlation is
|
|
R is negative when there is...
|
An inverse correlation
|
|
What do correlations of 1 and 0 mean?
|
1 = perfect reliability
0 = no reliability |
|
The more residuals you have, the weaker the correlation, and R is closer to...
|
- R is closer to 0
|
|
If all residuals are close to the line of best fit, then R is closer to...
|
- R is closer to 1
|
|
Random Error
|
- No idea where it came from
- Associated with reliability |
|
Systematic Error
|
- Consistently and always biased in a certain direction (more of a solvable problem)
- Associated with validity |
|
The amount of residuals and how weak the correlation is is directly related to...
|
- Directly related to random error and how much of it there is
|
|
Requirements for Validity Review:
|
- Agreed-upon DEFINITION
- Measurement methods SPECIFICALLY collect info. about events that fit that definition - Measurement methods COMPLETELY assess the whole concept described by that definition - IF there is another measurement method for the same concept that is known to be valid, the new method should have both a strong correlation with the GOLD STANDARD method and little bias (little/no difference in paired T-test) |
|
Paired T-test
|
- Tests for an overall difference in the average between 2 repeated measurements
- For each individual, calculate measure 1 minus measure 2 (subtract the difference) - If there is no difference (Null), then the distribution of the differences should be centered about zero ---> Difference from zero is bias ---> Greater variation/spread of these differences is random error |
|
Predictive Validity
|
- Does the measure help us understand the real world?
- Measure something at time 1 and then measure something else at time 2 and see if they correlate in the expected direction - 2 totally different measurement techniques with 2 different scales; so, do not need slope=1 and int.=0 |
|
Discriminant Validity
|
- Wanting the measure to be different from something else, to show a contrast between them
- Looking for a low or zero correlation - Want to show that they are separated |