Research Misc. Flashcards
Describe Face validity
The extent that a tool measures what it is supposed to from a superficial perspective
(looking at yourself in the face, superficial without a deep dive)
Describe concurrent Validity
Comparing a tool to the gold standard (comparing to another current tool)
Describe construct validity
How much does a tool measure an abstract concept
(construct=abstract ideas, does it match with the theory)
Content validity is what
How much does the tool measure the specific subject matter
(e.g does algebra test have algebra or geometry questions, getting deeper into the content)
Validity vs Reliability
Validity: How close to truth. This is why I say people ae not valid. They are far from truth
Reliability: How consistent are you. I say kyrie isn’t reliable b/c them 9 point games
What is a quasi-experimental design
Research that seeks to establish causality without manipulation (no randomization)
compared to an experimental design when the independent variable is manipulated (difficult in PT because can’t hide intervention)
What is the null hypothesis
Means No relationship between groups.
We want to DISPROVE THIS, meaning prove that there is relationship
Independent vs dependent
Input vs output/result
What is an alternate hypothesis
When there is a relationship between groups (typically what we want to be correct)
Normal curve
Would have mean, median and mode at same point
Intra-rater vs inter-rater
Intra-rater in how likely is same person going to get consistent same results
vs
inter-rater is “between” diff researchers how consistent
Internal vs external validity
Internal validity - any other reasons for outcomes (confounding factors)
External validity - Is it generalizable
Nominal scale
Named, no significance besides the given name
Ordinal scale
E.g is Borg RPE scale. Subjective scale with order.
MMT grading. Cant tell exact difference just know order/direction
Interval scale
Distance between each is meaningful. E.g is temperature or PH
Ratio scale
An absolute zero.
E.g is weight no such thing as negative.
T-test
The degree to which TWO groups of data are different. (Means divided by variability of groups)
=Comparing the means of 2 groups to see how diff they are
*z test would be the same only unknown
If P<0.05
Reject Null Hypothesis
This means there is a relationship which is what we are looking for
If P>0.05
Accept Null hypothesis, there is no relation or change that comes due to independent variable
Chi-squared test (X^2 test)
Assesses difference between observed and expected values to determine how they are or are not related.
Compares two different categorical variable and if they are related
Pearson Correlation Coefficient
Positive correlation: x and y increases. Notice line goes up from left to R
Negative correlation: y increases as x increases
Notice line goes down from R to L
No correlation: random points = 0.0
-1 to 1
closer to -1 or 1 means stronger and more linear in either direction
Regression analysis
using correlation coefficients to predict values
What percent of people fall within 2 SD
95
Standard deviation
Determining how far off value is from mean
T-score for bone uses this
-1 to 1 is healthy
-1 to -2.5 is osteopenia
less than -2.5 is osteoporosis
Use 50 percent to assist you on SD math questions
what percent of people fall within 1 SD in normal curve
68
What percent of people fall within 3 SD
99.7 (more spread out)
ANOVA Testing
Analyzing 3 more data sets
Sensitivity
% if pt Correctly identified positives
What percent of true positives were detected
SnOUT
High SENsitive test NEGative result can be confidently ruled OUT
because low amount of false negaives
Good for screening
Specificity
% of pt Correctly identified negatives
What percent of true negatives were detected
SpIN
High SPecificity and POSitive result means can be confident that positive rules IN
b/c positive test
Good for confirming b/c low amount of fale positives
Likelihood ratio
Determines how much a diagnostic test increases or decreases likelihood of having condition
+LR
Indicates a positive test rules in (higher the number the greater the likelihood that it rules in) starting at 1
- LR
Indicates a negative test rules out (lower the number the greater the likelihood that it rules out) starting at 1and decreased towards 0. the smaller the greater likelihood
Z-score
Strictly used SD to describe how far from mean (compared to population NOT norms )
T score
Compares individual to normative value NOT population (e.g healthy individual) with SD
of sample pop
e.g for bone density T score
Alpha level
Determined by researchers…what is the set level for probability of rejecting null hypothesis when it is actually true. Usually set at <0.05
P level
The probability of that results are lucky/due to chance if hypothesis is true. meaning typically less than 5% chance that it null hypothesis is true due to luck
If p value is less than alpha level, you rject the null hypothesis
List levels of research from most reliable/least bias to least reliable/most bias
- Systematic Review (:
- RCT
- Cohort study
- Case control
- Cross sectional/survey
- Case report
7 Mechanistic studies - Editorial/ expert opinion
Type 1 error does what
Rejects the null hypothesis when null hypothesis is true (false positive)
e.g Telling someone they have ACL tear when they do not
This is false positive b/c your fall is about what is true
Saying there is a relatinship or postive test when there isnt
Type 2 error does what
Accepts null hypothesis when it is false (not true)
False negative
False either way cause its an error. This is false negative b/c its a double negative
e.g Telling someone they do not have an ACL year hen they do
Saying there isnt a relationship when there is
MCD vs MCID
MCID: Smallest change that would be important to patient
Thats why its minimal CLINICALLY important difference
MCD: smallest amount of change that can be detected and is reliable
50 meter change is MCID for 6MWT
NNT
Number needed to treat to prevent one bad outcome
The ideal number is 1 (meaning each treatment is effective at preventing a negative outcome)
Case studies
Are for describing a single phenonomenon
Types of studies
A prospective cohort would require the investigators to follow the patients until the outcome. Because the information was collected from medical records, and the outcome has already occurred, the stem does not describe a prospective cohort study. (p. 287)
2. The patients in this study have already received the exposure (intervention) and have already experienced the outcome. Therefore, this is a retrospective study. (p. 288)
3. A cross-sectional study is used to assess exposure and outcomes at a single point in time. This study abstracted data over a period of time and, therefore, cannot be a cross-sectional study. (p. 280)
4. A case control study classifies people based on whether they had an outcome of interest and then looks retrospectively at different exposures. Because the sample in this study is compared on the basis of exposure (intervention), this study cannot be a case control study. (pp. 282-283)
