Term
|
Definition
data that can be divided into groups
-race, sex, age education status |
|
|
Term
|
Definition
divided into qualitative categories or groups
-male/female, black/white, urban/surburban/rural, red/green
-no implication of order or ratio |
|
|
Term
|
Definition
| -nominal data that falls into only two groups |
|
|
Term
|
Definition
-can be placed into meaningful order
-i.e. 1st, 2nd, 3rd
-no info about size of interval
-can't determine difference btwn 1st and 2nd, 2nd and 3rd etc |
|
|
Term
|
Definition
-can be placed in meaningful order
-have meaningful intervals btwn items
-i.e. Celsius scale |
|
|
Term
|
Definition
-can be placed in meaningful order
-have meaningful intervals btwn items
-have an absolute zero--> meaningful ratios do exist (i.e. weight, time, bp, pulse rate) |
|
|
Term
| What is normal (Gaussian) distribution? |
|
Definition
-symmetrical, bell-shaped curve
-mean=median=mode |
|
|
Term
| For Gaussian distributions, what data proportions are contained within +/- 1 SD? |
|
Definition
|
|
Term
| For Gaussian distributions, what data proportions are contained within +/- 2 SD? |
|
Definition
|
|
Term
| For Gaussian distributions, what data proportions are contained within +/- 3 SD? |
|
Definition
|
|
Term
|
Definition
| -how many SDs a point lies above or below the mean of distribution= converted into a % |
|
|
Term
| What is the equation to calculate a z score? |
|
Definition
|
|
Term
|
Definition
| -area under the curve from Z score to end of the curve |
|
|
Term
| What is the actual number of SDs for 95% of data? |
|
Definition
-area under the curve on one side= 0.025
-Z score= 1.96 |
|
|
Term
| What is the actual number of SDs for 90% of data? |
|
Definition
-area under the curve for .05
-1.65 |
|
|
Term
| What is a positively (right) skewed non-Gaussian distribution? |
|
Definition
-Mean> median> mode
-elongated tail at the right; more data in the right tail than would be expected in a normal distribution |
|
|
Term
| What is a negatively (left) skewed non-Gaussian distribution? |
|
Definition
-Mean< median< mode
-elongated tail at the left; more data in the left tail than would be expected in a normal distribution |
|
|
Term
| Discrete vs. continuous data |
|
Definition
-discrete- can only take certain values and none in between (ex. # of patients in a hospital census, # of syringes used in one day)
-continuous- may take any value (typically between certain limits); most biomedical variables are continuous (i.e. weight, height, age, BP); however, reporting the vales reduces them to a discrete variable |
|
|
Term
| What is the level of significance, alpha? |
|
Definition
-probability level at which it is decided that the null hypothesis is incorrect
-< 0.05= difference between the groups is significant- null hypothesis is incorrect
->0.05= difference between the groups isn't significant- null hypothesis is correct. |
|
|
Term
| What is statistical significance? |
|
Definition
-probability that the events were unlikely to have occurred by chance (i.e. p< 0.05= likelihood of results having occured by chance is 0.05 or less)
-does NOT make result truly significant |
|
|
Term
| When is the Chi square (x^2) test used? |
|
Definition
-used for testing hypotheses about nominal scale data
-Test of proportions- whether the proportions of observations falling in different categories differ significantly from proportion that would be expected by chance.
-compares percents or proportions (not mean values) |
|
|
Term
| When is the Student's t test used? |
|
Definition
-T score= estimated standard error used to find a statistic; must be used when making inferences about means that are based on estimates of population parameters rather than on population parameters themselves
- 1+ comparisons (i.e. 1+ groups being compared) |
|
|
Term
|
Definition
-Total variability- variability resulting from known differences btwn groups; ordinary random variability within each group that is to be expected in any set of data caused by sampling error, individual differences between patients, etc
-Compare variance- F ratio= (variance between groups)/ (variance w/in group)
-1 comparison (ie. between two sample means, or between sample man and hypothesized population mean. |
|
|
Term
| When do you use Wilcoxon rank-sum? |
|
Definition
-used when comparing two related samples, matched samples, or repeated measurements on a single sample to assess whether their population means rank differ
-used as alternative to Student's T test when population can't be assumed to be normally distributed
-interval scale data |
|
|
Term
| When do you use Manten-Haenzel? |
|
Definition
| used for repeated tests of independence |
|
|
Term
| When do you use Mann-Whitney? |
|
Definition
-used to assess wheter 1 or 2 samples of independent observation tends to have larger values than the other.
-ordinal data
-distributions sufficiently far from normal and for sufficiently large samples |
|
|
Term
| When do you use Kruskal-Wallis? |
|
Definition
-used when there is 1 nominal variable and 1 measurement variable, and measurement variable doesn't meet assumption of ANOVA
-NO assumptions about normality
-Ranke data |
|
|
Term
| What is a "non-inferiority" trial? |
|
Definition
| -primary objective is response to investigational product is not clinically inferior to a comparative agent (active or placebo control) |
|
|
Term
| What is a "1 tailed analysis"? |
|
Definition
-alternative hypothesis is directional (aka, specifies that the population mean lies in a particular direction with respect to null hypothesis.
-more powerful than two tailed analysis
-used w/ "non-inferiority" clinical trials |
|
|
Term
| What is a "2 tailed analysis"? |
|
Definition
| -alternative hypothesis is nondirectional (aka, population mean not equal to X, but didn't specify whether above or below) |
|
|
Term
| Do "non-inferiority" trials primarily use 1-tailed or 2-tailed analysis? |
|
Definition
|
|
Term
|
Definition
-gold standard for clinical trial
-tests efficacy of various types of intervention within a patient population
-subjects randomly allocated to receive one or the other treatment |
|
|
Term
|
Definition
-focus on factors related to disease development
-aka "which factors can cause disease?"
-some people exposed to risk factor for disease
-"follow-up" study or "longitudinal" study |
|
|
Term
|
Definition
-compares people with disease and those without
-looks to identify possible independent variables that cause disease
-aka "how did I get this disease and the other person didn't?" |
|
|
Term
| What are some advantages to a case-control study? |
|
Definition
-performed quickly and cheaply
-require comparatively few subjects
-allow multiple potential causes of disease |
|
|
Term
| What are some disadvantages to a case-control study? |
|
Definition
-bias about history based on presence of disease
-undiagnosed/asymptomatic cases missed
-comparable control group difficult to assemble
-control group is based on researcher's judgment
-cannot determine rate or risk of disease
-no cause-effect relationship |
|
|
Term
|
Definition
-surveys entire population
-assesses proportion of people with certain disease
-aka "how many people in Kentucky have disease X?" |
|
|
Term
| Disadvantages of cross-sectional survey |
|
Definition
-likely to over-represent chronic disease and under-represent acute disease
-may be unstable for acute disease (few people suffering from it at the time of survey)
-people with certain disease may leave community, be institutionalized |
|
|
Term
|
Definition
-ratio of the odds that a case was exposed: odds that control was exposed
-OR = (case exposed to risk)/(control exposed to risk)
-must be used when analyzing case-control data |
|
|
Term
|
Definition
-relative risk
-(incidence of disease among people exposed)/(incidence of disease among people not exposed) |
|
|
Term
|
Definition
|
|
Term
|
Definition
-difference between upper and lower confidence limits
-mean +/- (z score*SD) |
|
|
Term
|
Definition
-problem with study design
-study consistently errs in a particular direction |
|
|
Term
|
Definition
| overestimation of survival duration among screen-detected cases (vs. those detected clinically) when survival is measured from diagnosis |
|
|
Term
|
Definition
| overestimation of survival duration among screening-detected cases caused by relative excess of slowly-progressing cases |
|
|
Term
|
Definition
exposure status measurement relies on patient report
-patients can be poor historians |
|
|
Term
|
Definition
| unintended systematic difference between study groups |
|
|
Term
|
Definition
variables that contribute differently and inextricably to the 2 groups
-this is why you need random assignment |
|
|
Term
|
Definition
observed association due to causal relationship
-measured via randomized clinical trial or cohort study
-can't be measured via observational study |
|
|
Term
|
Definition
"false positive"
-accepting the alternative (or experimental) hypothesis when it's wrong
-probability Type 1 error will be made = a = p |
|
|
Term
|
Definition
"false negative"
-rejecting the alternative (or experimental) hypothesis when it's true
-probability Type 2 error will be made = B |
|
|
Term
|
Definition
probability that false null hypothesis will be rejected
- 1-B |
|
|
Term
| How do you increase the power analysis? |
|
Definition
-increase sample size
-increase a
-increase size of difference between sample mean and hypothesized population mean
-decrease sampling error |
|
|
Term
| If p > 0.05, what should you look for |
|
Definition
| -see if they did a power analysis to prevent "type 2 error" |
|
|
Term
| How do you calculate "number needed to treat" in "positive" clinical trials? |
|
Definition
| NNT= 1/ARR = 100/[(risk with usual care)-(risk with new tx)] |
|
|
Term
|
Definition
-ability to detect people who DO have disease
-rule OUT people who don't (snOUT)
= [TP / (TP + FN)] x 100 |
|
|
Term
|
Definition
-ability to detect people who do not have disease
-rule IN people who do (spIN)
= [TN / (TN + FP)] x 100 |
|
|
Term
|
Definition
-number of people tested positive when disease is actually absent
-Type 1 error |
|
|
Term
|
Definition
-numer of people tested negative when disease is present
-Type 2 error |
|
|
Term
| Positive predictive value |
|
Definition
-proportion of positive results that are truly positive
PPV = TP / (TP + FP) |
|
|
