Reading...
Play button
Play button
Use LEFT and RIGHT arrow keys to navigate between flashcards;
Use UP and DOWN arrow keys to flip the card;
H to show hint;
A reads text to speech;
32 Cards in this Set
- Front
- Back
|
What is a proportion?
|
Number in study with a factor
DM / (DM + NDM) |
|
|
Relative Risk?
|
For people with disease, what is ratio of people with a factor over people with it?
(DM&C / DM) / (NDM&C / NDM) Increased number of times |
|
|
Attributable Risk?
|
(DM&C / DM) ---- (NDM&C / NDM)
Increased percentage |
|
|
Relative Risk Reduction?
|
1 - RR = % reduced if ______
|
|
|
Absolute Risk Reduction?
NTT? |
% without risk ---- % with risk
NNT = 1/ARR |
|
|
SnNout?
|
sens test = neg rules dis out
TESTS PRESENCE OF DIS TP / All dis outcomes (TP + FN) |
|
|
SpPin?
|
spec test = pos rules dis in
TESTS ABSENCE OF DIS TN / All no dis outcomes (TN + FP) |
|
|
Odds Ratio?
|
NO incid/prev
ALZ Exposed & DM/ NDM ______________________ Control & DM / NDM (107/400) / (293/400) ________________ (60/400) / (340/400) |
|
|
Reliability vs. Validity?
|
Reliable = precision CV
Valid = accurate |
|
|
Pearson?
Spearman? Point Bis? Phi coeff? % var? |
P = continuous
S = ranked PB = ordered Phi = magnitude of correl % V = % one factor variance explained by a another |
|
|
+ PV?
-PV? BAYES CORRECTS PV FOR PREV |
+ = TP / ALL P (FP)
- = TN / ALL N (FN) if test is pos or neg what is percent likelihood that they actually have/dont have? |
|
|
LR?
|
SAME AS PRED VALUE, if disease present, what is likelihood of pos test?
LR + = SN / 1-SP LR -- = 1-SN / SP |
|
|
Confidence Interval?
|
MEAN +/-- (Z * SEM)
95% confident that your values include pop mean (you are right 95% of time) low CI = lower precision (68%? +/- 1 SD) Difference btwn sample and pop multiplied by value for +/- 2 SD for pop based on SAMPLE! |
|
|
Add P's?
|
probability of two mutually exclusive events occur
blood A or AB? |
|
|
Mult P's?
|
probability of two non mutually exclusive events occur (can both happen at same time)
Blood A and Female? |
|
|
alpha?
|
probability a difference is due to chance
p<.05 is good and not due to chance |
|
|
If a one tailed alpha=.05 what does two tailed have for it to be the same?
|
.025
|
|
|
POWER?
|
Correctly accepting or rejecting null
depends on N and alpha |
|
|
Risk of type two error increases as alpha becomes?
|
more stringent, N matters too
Saying no difference when there really is! |
|
|
type one error depends on ___
type two error deponds on____ |
1 = alpha
2 = N |
|
|
t = ?
|
mean / (sq root of var/n)
WHETHER MEANS OF TWO GROUPS ARE DIFFERENT (SAME OR DIFF POP) as n increases, more like pop dist N > 30, then t same as z if t = 2 and z = 1.96, then you are outside the 95% CI so THERE IS A DIFFERENCE AND REJECT NULL |
|
|
Cohen's D
|
diff between the means /
sq root of variance sums EFFECT > .8 IS LARGE! significant |
|
|
SS?
|
sum of differences of each value from mean
|
|
|
if increase ANOVAS/comparisons, what happens to error?
|
alpha compounds, ERROR INCREASES
|
|
|
Chi squared used for?
|
comparing frequencies in two or more groups
U.S. expected over hosp observed blood types |
|
|
Z score?
|
pt number - mean / SD
= number of stand deviations Normalize: mean +/- (SD * Z) to make normal distribution |
|
|
Central limit theorem?
|
keep drawing samples and make distribution w/averages of samples
will form normal dist |
|
|
+/- 1 SD?
+/- 2 SD? +/- 3 SD? |
68%
95% 99.7% an outlier is more than 6 SD |
|
|
VARIANCE?
|
sum of the squared deviations / N
|
|
|
Stand dev?
|
sq root of variance
|
|
|
CV?
|
(SD / mean) * 100
true value lies w/in certain percentage of the mean measure of precision/consistency LOWER CV = HIGHER PRECISION |
|
|
SEM?
|
SD sample / sq root N
HOW CLOSE SAMPLE IS TO POP diff btwn sample and pop means larger n has smaller variability POP MEAN SHOULD BE W/IN 2 SD OF SAMPLE MEAN |
