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;
45 Cards in this Set
- Front
- Back
|
Random Error |
- Due to chance Should go to zero over time - due to observer or subject variability, instrument wear
|
|
|
Enhance precision & accuracy by |
1. Standardize measurement methods 2 train and certify observer 3 refine and automate instruments 4. Repeat measurement. |
|
|
Systematic error |
Error due to bias |
|
|
Validity |
How well measurement represents variable of interest |
|
|
Bias |
Anything other than the experimental variable that will increase the probability of an outcome |
|
|
Reliability |
Consistency of measurement between and within observers - Interrater: two or more observers seeing the same thing |
|
|
Reliability |
Consistency of measurement between and within observers - Interrater: two or more observers seeing the same thing |
|
|
Hawthorne Effect |
Knowing you are in a study influences your response/ behavior |
|
|
Rosenthal effect |
Effect of investigator on response (subject might work harder for a good doctor) |
|
|
Type I error |
Rejecting the null when you should not |
|
|
Type II error |
Accepting the null when you should not |
|
|
p-value |
how compatible data is with the null, if p is low then null must go |
|
|
Confidence intervals |
- amount of uncertainty associated w/ sample, "95% confident that the mean would fall between ____ & _____" - if null is contained within intervals, then accept the null = not stat sig. |
|
|
Nominal data |
categories - labels |
|
|
Ordinal data |
think of scales - pain scales, 1st 2nd 3rd |
|
|
Interval "continuous" data |
interval between values are even - temperature |
|
|
Independent t-test |
samples are independent of each other |
|
|
t-test requirements |
1. must have two samples that are representative of group 2. categorical independent variable 3. continuous dependent variable 4. each group is normally distributed |
|
|
Cross-Sectional Study Advantages |
Think survey 1. short duration, no loss to followup 2. good first step for cohort studies 3. Yields prevalence of multiple predictors + outcomes |
|
|
Cross-Sectional Disadvantages |
1. Doesn't establish sequence of events 2. Not feasible for rare cases 3. Does not yield incidence |
|
|
Cohort Study General Advantages |
1. Establishes sequence of events 2. Assess multiple predictors & outcomes 3. # of outcomes may grow overtime 4. yields incidence, relative risk, excess risk |
|
|
Cohort Study General Disadvantage |
1. Often requires large sample 2. Less feasible for rare outcomes 3. Loss to follow-up |
|
|
Prospective Cohort Advantages (Outcomes occur in future Exposure occurs in present) |
1. More control over subject and measurement selection 2. May involve intervention 3. Using inclusion & exclusion, you have good control over sample 4. Can calculate incidence |
|
|
Prospective Cohort Disadvantages |
1. Follow up can be lengthy 2. Can be expensive 3. Loss to follow-up 4. Inefficient for studying rare outcomes |
|
|
Retrospective Cohort Advantages |
Advantages 1. Follow-up is in the past 2. Inexpensive Disadvantages: Less control over subject selection & measurement |
|
|
Multiple Cohorts |
Advantage: Useful when distinct cohort has rare exposure Disadvantage: Bias & confounding from sampling distinct populations |
|
|
Case Control Advantages (Exposure in Past) Retrospective in nature |
1. People already have disease, so it's useful for rare outcomes 2. Short duration, small sample 3. Inexpensive |
|
|
Case Control Disadvantages |
1. **Bias & confounding from sampling two populations 2. One one outcome studied at a time 3. Rely on people to recall exposures 4. Selection bias (lack of control over controls) |
|
|
Randomized Control Trial Advantages |
- Randomization prevents selection bias by uniformly distributing confounding variable - Randomization occurs after inclusion & exclusion criteria employed - Decision of placebo or comparison drug |
|
|
Risk Ratio COHORT Studies |
- ratio of # who developed outcome to # at risk - a/(a+b) divided by c/(c+d) - if RR > 1 then there is that many times the risk of developing the disease |
|
|
Odds Ratio CASE CONTROL |
- Used in Case Control - ratio of # of who developed outcome to # of those who did not - ad/bc - If OR > 1 then greater chance of being exposed - If null (1) is within the confidence interval you must ACCEPT the NULL |
|
|
How to control bias: |
1. Good research design 2. Blinding (masking) 3. Clear instructions to subjects 4. Treating everyone the same with exception to intervention |
|
|
Interrater |
two or more observes seeing the same thing |
|
|
Selection (Threat to Internal Validity) |
- subjects are selected in a way that differences may already exist before treatment is applied |
|
|
History (Threat to Internal Validity) |
events that occur during the time intervals between treatments |
|
|
Instrumentation (Threat to Internal Validity) |
Inconsistencies in the conditions, pretest/posttest not equivalent, or scorer that creates an illusory change in performance |
|
|
Testing (Threat to Internal Validity) |
Exposure to pretest influences performance on posttest |
|
|
Experimental Mortality (Threat to Internal Validity) |
Subject attrition may bias the results |
|
|
Maturation (Threat to Internal Validity) |
changes that occur within the subject during treatment: growing older, hungrier, etc |
|
|
Interaction of selection & maturation (Threat to Internal Validity) |
interaction between the selection of groups and maturation may lead us to believe that the treatment caused the effect - selection criteria makes it more likely for dropout |
|
|
Statistical Regression (Threat to Internal Validity) |
High & low values will naturally regress toward the mean |
|
|
Interaction effect of testing (Threat to external validity) |
exposure to pretest my sensitize the subject to the variable. - the general public has no pretest, so how can you generalize the results to them |
|
|
Interaction effect of selection bias & experimental variable (threat to external validity) |
- the effect of treatment may interact with certain characteristics within the experimental groups |
|
|
Interaction of experimental arrangements (threat to external validity) |
- being exposed to experimental conditions influences the results - hard to observe without effecting that which is being observed - Hawthorne Effect |
|
|
Interaction of multiple treatments (threat to external validity) |
- effects of prior treatments are not erasable |
