STATISTICS Flashcards
variable that is presumed to cause a change in another variable
independent variable
categorical variable
RELIGION
interval data
corresponds to a NUMBER
1) DISCRETE
2) CONTINUOUS
opposite of categorical data
discrete?
even numbers ONLY (aka number of toes)
continuous?
constant scale….lots of decimal
categorical data types?
1) dichotomous
- male/female
- dead/alive
2) nominal
- no ranking; blue eyes
3) Ordinal
- ranked data, but NO consistent scale (ex - ASA class, but hard to quantify difference between them)
cannot use mean/median for which kind of data?
categorical data (eye color)
weakness for Mean?
outliers
symmetrical data
median useful for what kind of data?
Skewed data
can’t use mode with what sort of data?
continuous (too many decimal points, won’t have any repeating)
variance?
average of SQUARED deviations from the mean
easier than variance?
-standard deviation
standard deviation relates to variance how?
square root of variance
how much within 1 standard deviation of mean?
68%
how much within TWO standard deviations of mean?
95%
three SD?
99.7%
Probability ASSUMES:
-normal distribution
99% confidence interval has ____ range of values than 95% CI
LARGER (more confidence, larger range)
null hypoth: defendant is innocent
alternative: defendant is guilty
rejecting the null
the defendant is guilty
TYPE 1 ERROR
false positive
convict when innocent
(you’re pregnant…as a male)
how type 1 error relate to Null?
reject a TRUE Null
type II error relate to null?
don’t reject false null
guilty but gets off
statistical measure of the strength of the evidence?
P-value
alpha is the probability of WHAT?
type 1 error
if alpha is .05, what’s that mean?
5/100 chance a given result occurred purely by chance
ways to DECREASE type 2 error?
- increase alpha
- increase sample size (MOST COMMON)
- decrease sample variability
- increase difference measured btwn the compared groups
not noticing an error when one’s there?
type 2 error
P-value is the likelihood of Type __ error?
Type 1
parametric data (categorical things that u can measure) , use THIS test?
T-test
(comparing two different things):
ONE-SAMPLE:
-comparing different things on the same object
TWO-SAMPLE (unpaired)
-comparing something, but on two different objects
ANOVA (analysis of variance) used for?
when many different tests performed, but would be too many samples and too much error
–puts all data into one number (F) and gives you one probability for the null hypothesis
non-parametric data is what?
NOT NORMAL DISTRIBUTION
-or ordinal/categorical (eye color, ASA class)
tests for non-parametric data?
- Mann-whitney U test
- Wilcoxon T-test
- Kruskall Wallis H-test
- Friedman x2 test
Chi-square tests for?
TWO things are independent
correlation analysis?
strength of relationship between variables (pos or neg)
Regression analysis?
mathematical equation to describe relationships of variables
R-square tells us?
Proportion of variability in Y accounted for by X
–aka, STRENGTH of relationship btw variables x&y
R-square values?
1 = perfect linear relationship 0 = no linear relationship
multiple regression used when?
describe multiple independent variables are related to SINGLE dependent variable
ANOVA
multiple variables, but CATEGORICAL data
ODDs ratio:
corresponds to prob of event occurring vs not occurring
1 - more likely in first group
RISK ratio
calculated same way as OR
diff btw OR and RR?
OR = no implication of temporality of association;
RR = risk of developing one condition IF you have exposure
OR uses what studies?
CASE CONTROL
RR uses what studies?
COHORT studies
PPV calc?
true pos test/all positive TESTS
NPV calc?
true neg test/all neg TESTS
sensitivity?
pos test with actual dz/total have dz
specificity?
neg test and no dz/total neg dz