Basic statistic (4)

Paired student t test 
 - compare 2 paired groups of patients with parametric data (normal distribution)
Wilcoxon test
 - compare 2 paired groups of patients with non parametric data (skewed distribution)


Unpaired student t test
 - compared 2 unpaired groups of patients with parametric data (normal distribution)
Mann-Whitney U test
 - compared 2 unpaired groups of patients with non parametric date (skewed distribution)

Chi-squared test
 - to compare percentages or proportions of categorical data (2 X 2 contingency table)

Basic statistic (3)

If a drug reduces the incidence of stroke from 10% to 5%, 

Relative risk reduction (RRR) is 50%

Absolute risk reduction (ARR) is 5%

OR

If the absolute risk of stroke is 10%,

the relative risk in treated group is 0.5,

then the absolute risk in treated group is 10 X 0.5 = 5%

the absolute risk reduction is 10% - 5% = 5%

Number needed to treat (NNT) is defined as number needed to treat to prevent 1 death

= 100% / ARR 

= 100% / 5%

= 20

Statistic (2)

Type 1 (alpha) error

- null hypothesis is falsely rejected (false positive)
- equals significance level of a test

Type 2 (beta) error

- null hypothesis is falsely accepted (false negative)

Power of a test (power of a test to detect a difference)

= 1 - beta

= 1 - (type 2 error)

Central limit theorem

- The random sampling distribution of any means would always tends to be normal, irrespective of the population distribution for which the sample were drawn. Hence, even if the population distribution is skewed or in any non normal distribution, the sample means would be normally distributed

- Mean of the random sampling distribution is equal to the mean of original population

Basic statistic (2)

OR = (38/25) / (69/129) = 2.84

Basic statistic (1)

Data
- Qualitative
   Nominal : ABO blood group, male/female, dead/alive
   Ordinal  : mild/moderate/severe
- Quantitative
   Numerical discrete : number of death per year
   Numerical continuous: age, weight, blood pressure


Sensitivity
- probability that a test will be positive when a patient has the condition (TP / TP+FN)


Specificity
- probability that a test will be negative when a patient does not have the condition (TN / TN+FP)
- if threshold for screening is increased, specificity will increase


Positive predictive value
- probability that if a test is positive, it is true (TP/ TP+FP)


Negative predictive value
-  probability that if a test is negative, it is true (TN / TN+FN)


   The higher prevalence will have greater PPV and lower NPV