Research Flashcards
What is an independent variable
the thing you can manipulate- like different types of interventions- manual versus ultrasound for example
What is a dependent variable
it depends on type of intervention- like pain levels in response to ultrasound/ manual
What is the p value
Probability that differences between groups occurred due to chance
Ex: if U/s reduced pain by 4 points and manual reduced by 5 points. Is difference due to chance or is manual more effective than U/S
0.09- 9% chance that difference is due to chance
Any p value lower than .05 is statistically significant= alpha level
What is the alpha level
The p value level set by researchers- threshold must be lower to be statistically significant
Describe Type I and Type II errors
Type 1 error- the researchers determined there was a statistical difference between groups when there wasn’t- AKA backing a loser
Type 2 error- researchers don’t find statistical difference between groups when there actually was
AKA missing a winner- often due to too few subjects in the study (increase number of subjects to remedy)
Effect sizes- what are values/ ranges to determine what is significant?
If manual is better at reducing pain- how much better? Example- Effect size is .6= moderate 0.8 and greater is large .5- .799- moderate .2- .499- small < .2 is trivial
Interrater Reliability
Several combinations will all get same results
Intrarater reliability
Same clinician get same result when performing same result multiple times on the same patient
Cohens kappa (small upper case k) What are the cutoffs?
0= no reliability/ no better than chance
< .4= poor reliability
.4-.6= fair reliability
.6-.75= good reliability
> .75= excellent
1- perfect reliability (result same when performed on same person every time)
Example- study examined interrater reliability for SIJ mobility tests for SI dysfunction- Gillets, standing forward flexion test- examined students versus practicing clinicians- students had cohens kappa of 0-.3= none to poor. Practicing clinicians had 0-.167- none to poor again
Likelihood ratio- describe positive and negative
Tell you what to do with + or - test results
Positive- tells you how much you should increase your suspicion based on positive test results
Negative- tells you how much you should decrease suspicion base don negative test results. Will be smaller than 1
positive > 10= large shift towards diagnosis 5-10- moderate shift in probability < 5 small shift 1= no change
Negative < 0.1 = large shift away from diagnosis 0.1- 0.2= moderate > 0.2= small shift 1= no change
Example- SIJ pain CPR + likelihood ratio of 4.16- how much would suspicion shift toward SIJ pain if cluster is += small, same CPR has - likelihood ratio of 0.12 so if cluster is negative. So moderate shift away. Laslets classification is better at ruing out than in
Placebo effect
treatment has no medical/ therapeutic value. The individual believes the treatment is real, so they improve.
ex: fake pill that looks real
High quality research should use placebo group whenever possible, comparing treatment to placebo to make sure theres benefit
Note- effective placebo controls need to look like real treatment
Nocebo effect
Negative counterpart to placebo- negative beliefs results in worse outcomes/ negative reactions from subjects.
Example- if its suggested/ implied to expect negative side effects then the % is higher that negative effects will be reported.
Example 2- previous experiences will increase nocebo effect if they’ve previously had a negative experience
Hawthorne effect
Western Electric Plant in Illinois to see what effect environmental changes would affect productivity- lighting was increased/ decreased and either way productivity improved but went back to baseline after research ended
Subjects who know they are being part of a research study will work harder
For example- as clinician if you do same treatment and arent getting good results- assume the research group worked harder and had more adherence in HEP
Observer effect
people work harder when they’re being watched, in healthcare, report more improvement in response to more attention
Make sure the time spent with clinician is same among groups
John Henry effect
Railroad worker- when he heard about steam drill, worked so hard to beat it, he died in process
Control groups perceives they are disadvantaged to experiment group and they work harder than they otherwise would have to overcome disadvantage
Example- control group seeks out other treatment or performs more self treatment on their own, which threatens validity
Prevent by blinding control group so they don’t feel disadvantaged
