Quiz 2 Flashcards
mode
most commonly occurring value in a set of data
median
number that is halfway into a set of data when values arranged from least to greatest
mean
average
what are the three measures of central tendency?
- mode
- median
- mean
what are the three measures of variation?
- frequency
- IQR (interquartile range)
- SD (standard deviation)
IQR
interquartile range, middle 50 percent of a distribution
what are the three different types of variables?
- independent
- dependent
- confounding
what are the two types of scales and their subtypes?
- categorical (nominal and ordinal)
2. continuous (interval, ratio)
independent variables
variable which is manipulated by researcher
dependent variables
variable of interested, measured or observed by the researcher (O from PICO question)
confounding variables
unobserved, unmeasured variable that unknowingly influences the dependent variable
nominal data
classification or label, w/ no implied order (e.g male and female)
ordinal data
categorical, rank order of observations (e.g. bad, medium, good)
interval data
units w/ equal interval, can be negative, not representing an absolute quantity (e.g. temperature)
ratio data
units w/ equal interval, measured from true zero (e.g. height, weight)
categorical data
nominal and ordinal scales, nonparametric, use a special set of stat tests
continuous data
interval and ratio, parametric (uses math), standard stat methods
positively skewed distribution
tail points in positive direction
negative skewed distribution
tail points in negative direction
when should frequency distributions be used?
small data sets, these represent the frequency of certain responses in a data set
grouped frequency distribution
frequency distribution by category (example was PAs by gender and age)
central limit theorem
assumes normal distribution if samples is of normal population and there is a large sample size
regression to mean
expected change from natural variation (example was BP in clinic)
range
difference between the highest and lowest scores
variance
deviation of scores
standard deviation
square root of variance
what are the two aspects of statistical inference?
- hypothesis testing
2. estimation (point estimate vs. interval estimate)
point estimate
providing the best estimate of the true population value for a statistic
interval estimate
quantifying our uncertainty about how close our point estimate is to the true pop. value
null hypothesis
H0: position that you wish to test
alternative hypothesis
H1: probably alternative proposition if H0 is rejected
type I error
determining statistically that a difference between two samples exist when in reality a population difference DOES NOT exist (H0 is true but we say it is false)
type II error
determining statistically that no difference exists between two samples when in reality a pop. difference DOES exist (H1 is true but we say H0 is true)
when would type II error occur?
H0 is false but you accept H0
when would type I error occur?
H0 is true but you reject H0
probability
likelihood that any one event will occur, given all possible outcomes (certain=1, impossible=0)
P value
probability of a value as large or larger than the observed, given that H0 is true, higher P value means a higher chance of type I error
discuss what happens if alpha level is larger (p
greater to find significant, but greater chance for type I error (10%)
discuss what happens if alpha level is smaller (p
less chance of type I error (1%), but harder to find significance
power
probability that a test will detect a difference when once actually exists; probability that a test will lead to rejection of null
what is a “good” power value?
power>=0.80 is good
what three factors can increase power?
- maximizing difference between groups
- increasing sample size
- reducing variablity
standard error
hypothetical quantity that indicates degree of variability among sample means
CI
confidence interval; estimate of the potential range of values which are likely to include true value in real pop.; way of determining the precision in our estimate
how do you calculate the 95% CI?
point estimate +- 1.96 (standard error)
how do you calculate the standard error?
STDEV/(square root of N)
case-control study
observational, retrospective; patients w/ outcome of interest and control patients w/o outcome of interest, comparing to determine past exposures
case-series study
observational; report on a series of patients w/ an outcome of interest
cohort study
observational, prospective; involves ID of two groups, one w/ exposure, one w/o, following them to see if they develop outcomes
meta-analysis
systematic review that uses quantitative methods to summarize results
RCT
randomized into an experimental group and control group
systematic review
article in which authors systematically searched for, appraised, and summarized all lit. for a specific subject
best study design for therapies
systematic review/ RCT
best study design for diagnosis
blinded comparison w/ gold standard diagnostic test or procedure
best study design for prognosis/etiology
cohort/case-control
concealed allocation
investigators unaware of subject’s group in RCT
stratified
participants divided into homogenous groups based on possible confounding variables
computerized
algorithm to randomize groups in RCT
crossover design w/ washout period
one group, gets two treatments w/ washout period in between
cross-sectional
snapshot in time, prevalence survey, community survey, large pop., can be used to compare communities w/ different risk factors
phase I
first in human, investigation of new drug in healthy humans, testing side effects
phase II
small sample, tests effectiveness of drug for indication, in affected pts.
phase III
larger population than phase II, gives additional info about effectiveness and safety
phase IIIb
testing in special population (example: geriatrics)
phase IV
post marketing phase, not all drugs require IV, only if deemed appropriate by FDA.
internal validity
integrity in the experimental design, performance of the experiment, effects are true for the study participants
external validity
appropriateness and extent by which results can be applied to non-study pop.
possible threats to internal validity
was the right instrument used? was it used correctly each time?
possible threats to external validity
was the study relevant to the research question? is the pop. similar to the one you will be treating?
bias
inaccuracy of measurements, this is a threat to internal validity
what are 6 types of bias?
- selection bias (picking sicker patients)
- observer bias (person taking measurements knows the tx group)
- participant bias (not following instructions)
- withdrawal bias (dropout rates should not exceed 20%)
- recall bias (example: study respondents not remembering things that were inconsequential)
- instrument bias
what is a cutoff for withdrawal bias?
dropout rate should not exceed 20%
precision
consistant readings
define accuracy
precise and unbiased (all arrows hitting the middle of the target consistently)
test-retest
consistent repeated measures
intra-rater
consistent single rater
inter-rater
consistent between different raters
intra-subject
consistency of a single subject
when should you use Mann Whitney?
non-parametric, independent, 2 groups
when should you use Kruskai Wallis ANOVA by Ranks?
non-parametric, independent, >2 groups
when should you use Sign Test or Wilcoxson Sign Ranks?
non-parametric, dependent, 2 groups
when should you use Friedman ANOVA by ranks?
non-parametric, dependent, >2 groups
when should you use t-test?
parametric, independent, 2 groups
when should you use ANOVA?
parametric, independent, >2 groups
when should you use paired t-test?
parametric, dependent, 2 groups
when should you use repeated measures ANOVA?
parametric, dependent, >2 groups
which conventional P value determines when you should reject H0?
reject H0 if P value