Week 2: Probability, Odds, OR, RR, RD, Chi-square, Central limit Theorem, Sampling Distribution, CI, St Error Flashcards
What is the difference between Probability and Odds?
Not the same!
Prob = 1 out of 5
Odds = prob of picking/prob of not picking=1 to 4
What measure of association is suitable to express the strength of the relation between one binary variable and one continuous variable
Pearson correlation
What measure of association is suitable to express the strength of the relation between two continuous variables?
OR, Pearson
What measure of association is most suitable to express the strength of the relation between two binary variables?
Pearson
For which design is the OR appropariate for 2 binary vaiables ?
a. A cohort study
b. A case-control study
c. An experiment with two conditions and a binary outcome
For which dsign is RR appropariate?
a. A cohort study
c. An experiment with two conditions and a binary outcome
Except CASE CONTROL, BECAUSE It is not meaningful to calculate the relative risk in a case-control study, because the ‘risk’ is defined by the researchers.
What frequencies are compared with the observed frequencies when the chi-square test statistics is obtained?
The frequencies that are expected if there is no relation between 2 variables
The chi-square test statistic for a contingency table can be calculated to examine the relation between
2 catego variables
Yes No
Restricted diet 15 89
Ad libitum diet 47 37
How would we calculate the Prob? The Risk for a r.d/yes?
The Odds r.d/yes?
The OR?
The RR?
The risk: 15/104 The odds: 15/89 OR: 1st odd/2nd odd RR: 1st risk/2nd risk Prob: 62/126 (sum ofyes/sum of no)
Yes No
Restricted diet 15 89
Ad libitum diet 47 37
Assume that there is no relation
15+89=104 47+37=84 104+84=188 104/188= 84/188=
What is the chi-square?
How is it calculated?
it tells us whether two variables are independent of one another.
x,y can have 2 or more levels but the groups formed by x are independent
if x is exposure, the exposed and non-exposed groups should be independent of one another
Chi-square relies on:
- having a large sample size
- it uses a theoretical probability distribution (Chi square distirbution)
- it assumes a large sample and a theoretical probability distribution
What is the Central Limit Theorem?
if the individuals sampled from the population are independent and if take a large sample size or if the distribution of the individuals is approx normal=> then the sampling distribution will be approx normal
When is the sampling distribution exactly normal?
When it is centered, symmetrical and unimodal.
When n→ ∞, because the bigger the n, the more the sampling distribution approaches a normal distribution.
The standard error (SE) of the average estimate depends on the sample size. What is the SE if the sample size is equal to n=100? What is the SE if the sample size is equal to n=10000?
SE (100)=s/√n=10/√100=3.16
SE (10000)=s/√n=10/√10000=0.1
Why should the standard error of the average estimate for n=100 be larger than for n=10000 (note: it is assumed σ is known and fixed in both samples)?
The standard error gets bigger for smaller sample sizes because standard error tells you how close your estimator is to the population parameter. So a sample of a bigger size, logically, would produce a closer estimate.
How does the sampling distribution look when the sample is equal to the whole population?
The mean will be exactly the same all the time. No variability. The value of the standard error is 0. The sampling distribution will look like ‘1 bar’ at the mean.
How does the sample distribution look like when the sample is equal to the whole population?
The sample distribution would look like the population distribution
How does the sampling distribution look like when the sample size is equal n=1?
The same as the observation mean. The distribution will have 1 baruniform distribution.
What is the difference between
- Sample Distribution
- Sampling Distribution
- Population Distribuion
