Statistics Flashcards
why are statistics important?
- some stats are predictive ( have to be reliable and valid )
- we are interested in the “application” of statistics
how can we represent DATA?
- mean (average)
- standard deviation
- standard error
standard error
- represents lots of individuals
- shows how much we can trust our mean (accuracy)
- the n variable rises as the number of people being measured increases
- used to generalize the mean to other cohorts or a population
- is dependent on SD and the sample size
standard deviation
- measure of variability within the cohort being assessed
- 2 x SD = 95% of the range of data
- used to describe the cohort of data
measuring variability
- includes the coefficient of variation which is Relative Standard Deviation
- RSD is the magnitude of the mean
- The SD is dependent in the mean.
- CV accounts for this
- CV% = SD / mean * 100
factors influencing variability
- biological
- technical
- testing
- environmental
- unknown
biological variability
- physiological and psychological fluctuations of the individual - circadian rhythms, mood, etc.
technical variability
precision and accuracy of the instruments
testing variability
instructions and manner of administering the test
( how you give tests/performances)
environmental variability
temperature and humidity
odds ratios and risk factors
- how something influences another outcome
- represents the effect of an “intervention” on a particular outcome
- both attempt to describe the same effect
- normalize the occurrence of an outcome in reference to a control group
- commonly used in medicine
validity
- accuracy and correctness
- does a test measure what it is supposed to measure?
- valid - invalid
reliability
- precision and repeatability
- consistency or replicability of a measurement
- type I error : not getting the outcome you should
- type II error: should be getting the outcome you want but don’t
logical or face value validity
- can be claimed when the measure appears to obviously assess the target variable/performance
- ex. balance test obviously measures balance
- weakest form of validity b/c it is difficult to quantify
- no statistical verification
- established by expert opinion or judges
content validity
- attempts to measure the desired parameter or a defined domain of context.
- applies to written tests or questionnaires
- often a table of specifications or diagrams are developed to act as a blueprint
- validity is established through published literature or curriculum content
- often no statistical verification is required
ex. visual rating scale for body comp
construct validity
- claimed when the measures permit inferences to be made about an underlying trait
- variable of interest is multi-factorial / multi-dimensional
- ## requires more complex statistical procedures such as factor analysis, multiple regression, etc.-
criterion validity
- the extent to which the results of a standard test can be compared to some criterion ( another test), or used to predict a practical outcome
- can be claimed when a test measure provides an outcome similar to a standard/criterion or previously validated test measure
- can ALSO be claimed when the measure taken, successfully predicts the criterion measure or gold standard
systematic error
situations that result in a unidirectional change in scores on repeated testing
random error
variability in a random manner, both increase and decrease test scores on repeated testing
testing reliability
- inter-rater
- intra-rater
- test-retest
inter-rater
- a measure of consistency used to evaluate the extent to which different judges agree in their assessment decisions.
ex. giving a score
intra-rater
comparison of two (or more) measures made by the same tester
- how good you are at getting an outcome/measuring
test-retest
- repeated testing on two or more occasions
- used to test the reliability of the technique (repeatability)
- getting similar results if the test is done twice
repeatability
- the same experimental tools
- the same observer
- the same measuring instrument, under the same conditions
- the same location
- repetition over a short period of time
- same objectives
intra-class correlation
interpreting the ICC
- less than 0.40 = poor
- btw 0.40 and 0.59 = fair
- btw 0.60 and 0.74 = good
- btw 0.75 and 1.00 = excellent
correlation
- describes the strength of the relationship between two variables of interest
- should be physiological basis fundamentally linking the variables of interest
- correlation IS NOT causation
ex. HR vs VO2 is valid because the VO2 equation has HR in it - does NOT describe the pattern of relationships
regression
- numerical relationship btw two variables
- how the variables line up with each other
- the simplest regression is line of best fit
multiple-linear regression
- many factors can be co-related, influencing the relationship of interest
bland-Altman test
- see PowerPoint slides
meta-analysis
- use to get at a big question
- pulls all the information together in a domain and normalizes a ton of information
- gold standard to defining where risks come from
- strict, defined process for conducting analysis
- pool data to provide a larger sample size
- highly regarding technique for interpreting variability / controversy in data
- validity of outcomes across studies
- basis for clinical practice guidelines
