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;
116 Cards in this Set
- Front
- Back
|
Descriptive statistics
|
describe, organize, summarize data Ex: avg. cholesterol values
|
|
|
Inferential statistics
|
make inferences based on data
Ex: sample cholesterol values from RANDOM # students |
|
|
a population
|
observations or measurements of ENTIRE subjects
|
|
|
a sample
|
a subset of the population-measurements of PARTIALLY selected subjects
|
|
|
another word for observation
|
element (X)
|
|
|
simple random sample
|
every element has equal prob. of being included
Ex: drawing fromm stack of 52 playing cards |
|
|
How do you eliminate biased samples?
|
True randomization...Stratified randome samples
|
|
|
stratified random sample
|
population divided into homogenous groups or strata (age, ethnicity, gender)
|
|
|
cluster sample
|
based on geographic areas
used when too expensive to draw simple random sample |
|
|
systemic sample
|
ex: select every 5th student
|
|
|
Facts about probability
|
cannot be negativ
expressed as decimals lie between 0 and 1 1-p=probabilty event wont occur |
|
|
When do you use the addition rule?
|
when the events are mutually exclusive
|
|
|
When do you use the multiplication rule?
|
When 2 or more events are independent and both could happen at the same time
|
|
|
Binomial distribution is normally...
|
to describe inheritence of genetic disease. The p with ONLY 2 possibilities
|
|
|
Nominal data
|
organized into qualitative groups: male/female
|
|
|
Ordinal data
|
data organized in ranking order
DOES NOT provide info on size of INTERVAL b/n data |
|
|
Interval data
|
data organized into meaningful order with meaningful intervals in b/n
DO NOT have ABSOLUTE ZERO EX: centrigrade scale |
|
|
Ratio data
|
data organized with meaningful intervals
DO have ABSOLUTE ZERO Ex: Kelvin scale, seconds, days, pulses/min |
|
|
cumulative frequency distribution
|
% of elements lying within and below each class interval
|
|
|
Ogive curve
|
S-shaped curve for cumulative frequency
|
|
|
What do centile ranks tell us?
|
% of observations that fall below any particular score
|
|
|
What is the shape of the normal distribution curve?
|
symmentrical bell shaped
Gaussian distribution |
|
|
How can you tell whether a skewed distribution is positive or negative?
|
by the location of the tail of the curve
|
|
|
Ex of central tendencies
|
mean, median, mode
|
|
|
Central tendency that occurs with the greatest frequency
|
Mode
|
|
|
What is the median when you have an odd # of elements?
|
middle number
|
|
|
What is the median when you have an even # of elements?
|
avg of two middle scores
|
|
|
Mean
|
avg
VERY sensitive to extreme scores |
|
|
3 measures of variability
|
range, variance, standard deviation
|
|
|
Difference b/n lowest and highest scores
|
range
|
|
|
How do you determine the deviation score?
|
difference between elements and the mean
* the sum of the deviation scores for all elements is 0 |
|
|
Can you use deviation scores to differentiate b/n 2 different normal distributions?
|
NO
|
|
|
What is the variance of a distribution
|
the mean of the squares of all the deviation scores.
1. find deviation scores 2. square each deviation score 3. obtain mean |
|
|
How do you find the standard deviation?
|
it is the square root of the variance.
|
|
|
What % of the distribution falls within +/- 1 s.d. of the mean?
|
68%
|
|
|
What % of the distribution falls within +/- 2 s.d. of the mean?
|
95%
|
|
|
What % of the distribution falls within +/- 3 s.d. of the mean?
|
99.7%
|
|
|
if element lies above the mean, it will have a ____ z score?
|
positive
|
|
|
if element lies below the mean, it will have a ____ z score?
|
negative
|
|
|
how to find z score
|
subtract mean from element and divide by s.d.
|
|
|
What is used for finding the probability that a random element will have a score above or below a mean of the distribution?
|
z score
|
|
|
What type of statistics makes conclusions about a population?
|
inferential
|
|
|
When plotted, what type of distribution will you see when plotting the random sampling distribution of means?
|
normal distribution even though the shape of the pop. distribution is rectangular
|
|
|
Are confidence limits 1 tail or 2 tails?
|
ALWAYS 2 tails b/c you need a max and min value
|
|
|
What is CI (confidence interval)?
|
the diff. b/n the upper and lower confidence limits
|
|
|
CI decreases in proportion to what?
|
the square root of the sample size (to halve the CI, the sample size must be increased 4 fold)
|
|
|
How is precision proportional to the sample size?
|
precision = square root of n
OR (precision)^2 = n |
|
|
As you increase sample size, what happens to the width of the CI?
|
width narrows
|
|
|
To double precision, how must you change the sample size?
|
multiply sample size by 4 to double precision
|
|
|
When something is precise, is it scattered or clustered?
|
clustered
|
|
|
When something is inaccurate, it it biased or unbiased?
|
biased
|
|
|
When are statistics precise?
|
when they are immune from random variation
|
|
|
precision is shown by the ____while accuracy is shown by the ______ b/n the mean of the random sampling distributions of means and the true pop. mean.
|
precision is shown by the width of the distributions (inversely)
accuracy is shown by the distance (inversely) |
|
|
Do you use the z score or t score when making inferences about means that are based on estimates of pop. parameters?
|
t score
|
|
|
For any given proportion of the distribution, is z constant? t?
|
z is constant while t is NOT constant b/c it depends on the size of the sample
|
|
|
When are z and t similar?
|
when the sample size is larger than 100
|
|
|
What do t values depend on?
|
degrees of freedom (df=n-1)
|
|
|
when do you use 1 tail? 2 tails?
|
1 tail is directional (improves, impairs)
2 tails is NON-directional (affects) |
|
|
What is alpha in hypothesis testing?
|
decision criterion/significan level: the point when the difference b/n the sample mean and the hypothesized population is due to chance or due to real effect
|
|
|
What is the Central Limit Theorem?
|
States that a mean of random sampling means will be very close to the true pop. mean
|
|
|
What is the conventional level of alpha?
|
0.05
|
|
|
If the probability that the sample mean could have come from the hypothesized pop is less than or equal to 0.05 what happens to the null hyp?
|
it is rejected
|
|
|
What is the range of acceptance?
|
the middle 95% distribution
|
|
|
What are the limits of the area of rejection defined by?
|
the critical values
|
|
|
What does it mean in terms of significance when the null hyp is rejected?
|
that the diff. b/n the sample mean and the hypothesized mean is statistically significant b/c it was rejected at the 0.05 level
(SIGNIFICANT) |
|
|
What does it mean in terms of significance when the null hyp is accepted?
|
that the diff. b/n the sample mean and the hyp population mean failed to reach statistical significance (NOT signifanct)
|
|
|
What does significant at p<0.05 mean?
|
an investigator can be 95% sure that the result was not obtained by chance. The diff was significant or real
|
|
|
What is a type I error?
|
rejecting the null when it is true (false negative)
|
|
|
What is a type I error also referred to as?
|
alpha error
|
|
|
What is a type II error?
|
accepting the alternative hyp when it is actually false (false postitive)
|
|
|
A type II error is also known as?
|
beta error
|
|
|
1-beta is what?
|
power of the test
|
|
|
what does power of the test tell us?
|
the probability that alt. hyp will be rejected
|
|
|
A test is reqd to a have a power of at least what to be acceptable?
|
0.8
|
|
|
What are the 2 errors of hypothesis testing?
|
alpha error: rejecting a null that is actually true
beta error: accepting a alt. hyp that is false |
|
|
what is the correlation b/n power of test and alpha?
|
power increases as alpha increases
|
|
|
are 1 or 2 tailed tests more powerful?
|
1 are more powerful b/c they are more strict
|
|
|
What is the chi square test for?
|
a test of proportions: testing hypothesis for nominal scale data
|
|
|
What's the difference b/n experimental studies and nonexperimental?
|
experimental: give drug to experimental group and compare effect w/ control group that hasn't taken drug
nonexperimental: observe the effect of drugs by comparing 2 groups who have ALREADY taken the drugs and who have not taken any--No ethical issues |
|
|
Purpose of clinical trials?
|
used to evaluate the effects of treatments and to isolate one factor by holding all other factors constant
|
|
|
What is the best method of assignment in RCCT trials?
|
randomization to reduce the selection bias so that any difference that appears b/n 2 groups at the end of the study can be attributed to TX
|
|
|
What is the diff. b/n double blind studies and single-blind studeis?
|
double-blind: neither subjects nor investigators know
single-blind:only the subject is unaware and isn't as effective b/c humans can't control their emotions |
|
|
In RCCT, how are the effects of confounding variables reduced?
|
by matching: 48 yr old male vs. 42 yr old male to see effect of drug (not 48 vs. 21) need to eliminate confounding variables: age and gender
|
|
|
Explain crossover designs.
|
a repeated measures design b/c the measurements are repeated w/n each patient at different times
Ex: patient A: drug 1st month, washout 2nd month, placebo 3rd month |
|
|
What study is the first method used to study a particular, new disease-also called exploratory studies?
|
Descriptive studies: most powerful method to study new disease
|
|
|
What studies fall under the category: nonexperimental studies?
|
descriptive studies
analytical studies cohort studies case-control studies case-series studies prevalence surveys |
|
|
What do analytical studies do?
|
aim to test hypothesis or to prove explanations about a disease after observing the particular disease
|
|
|
What happens in a cohort study?
|
group of young, healthy people selected and observed for an extended period (15+yrs) and followed forward from a particular point in time
Cohort= prospective |
|
|
Advantages to cohort study?
|
no ethical problems
lifestyle and health managment are the choice of individuals not investigators establish absolute risk NO bias |
|
|
Disadvantages to cohort study?
|
time-consuming
expensive impractical for RARE diseases |
|
|
Describe a case control study.
|
comparison is made b/n the cases (who have disease) and the controls (who do not)
Retrospective b/c they start w/ the outcome and then look back into the past |
|
|
Advantages of case-control study?
|
quick and cheap
important for RARE diseases need few subjects can establish multiple potential causes |
|
|
Disadvantages of case-control study?
|
highest degree of recall bias
loss in info of risk factors if people die cannot prove a cause-effect relationship |
|
|
Describe case-series studies.
|
Reports or presentations of a disease.
NO Controls DO NOT follow up used to present new info about patients with RARE disease and develop new Ho. |
|
|
Describe prevalence surveys "community surveys"
|
surveys of a WHOLE population
SINGLE examination of pop at a PARTICULAR point in time DO NOT follow up AKA: cross-sectional studies` |
|
|
Advantages of prevalence surveys?
|
b/c of community study, info on wide range of disease and characteristics can be gathered and used for hypothesis
|
|
|
Disadvantages of prevalence study?
|
not usable for ACUTE diseases
loss of subjects leaving from the community |
|
|
Eqn for incidence rate
|
# new cases of disease/total # people at risk x unit time
|
|
|
eqn for prevalance rate
|
# new cases with disease/total # people at risk at a PARTICULAR time
|
|
|
when can use prevalance rate?
|
for stable chronic condition: hypertension, diabetes
NOT for ACUTE DISEASE: appendicitis, pulm embolism |
|
|
eqn for prevalence?
|
annual incidence rate x average duration (yrs)
|
|
|
eqn for mortality rate?
|
total # deaths/total # of people at risk x time
|
|
|
What happens to prevalence if you increase incidence?
|
bathtub: increase incidence increase prevalence
|
|
|
eqn for case fatality rate?
|
# deaths due to disease/total # cases
DOES NOT depend on time |
|
|
eqn for attack rate?
|
# people contracting disease/ total # people at risk
(at 1 time, 1 incidence) |
|
|
absolute risk =
|
incidence rate of disease
|
|
|
Which study is the best for determining the effect of a risk factor?
|
cohort study: identify the risk factor of the disease
|
|
|
relative risk =
|
incidence among those EXPOSED to risk factor/incidence amoun those not exposed to risk factor
|
|
|
absolute risk reduction:
|
control incidence - experimental incidence
|
|
|
NNT=
|
1 life/absolute risk reduction
|
|
|
cost estimate eqn=
|
cost per month x months x NNT
|
|
|
attributable risk =
|
incidence exposed - incidence non-exposed
|
|
|
What kind of study will be used to find the odds ratio?
|
case control
|
|
|
eqn for odds ratio?
|
A x D / B x C
|
|
|
What does it mean when the odds ratio = 1
|
risk factor is NOT related to disease
|
|
|
What does it mean when the odds ratio is LESS than 1?
|
the risk factor may be a protective factor against disease
|
