Stata Lecture 4 Flashcards
How is proportion calculated for a single group?
(Number of participants in category)/(Total number of participants)
How is proportion represented?
Number between 0 and 1
What are the assumptions for calculating confidence interval?
That the data is normally distributed
There are at least 5 observations in each category
How is the odds of having the disease calculated for a single group?
(Number of participants in disease category)/(Number of participants in other category)
OR
(Proportion of participants in disease category)/(Proportion of participants in other category)
What is the relationship between proportion and odds?
The greater the proportion of people in the disease category, the greater the odds the you will have the disease
If proportion = 1, the odds = infinity
What are the two methods of summarising a binary variable for a single group?
Proportion and Odds
What would you use to compare groups in a case-control study?
Odds
What is the exposure variable vs the outcome variable?
Exposure variable = intervention/control
Outcome variable = measured
How is the risk difference calculated between two independent groups?
(Proportion with disease in intervention group) - (proportion with disease in control group)
How is the risk ratio calculated between two independent groups?
(Proportion with disease in intervention group)/(Proportion with disease in control group)
How is the odds ratio calculated between two independent groups?
(Odds of having disease in one group)/(Odds of having disease in the other group)
How to interpret the ‘risk difference’ result?
Risk difference of 0 = both groups equally likely to have disease
Risk difference > 0 = First group more likely to have disease
Risk difference < 0 = second group more likely to have disease
What is NNT?
NNT = Number needed to treat
Calculates effectiveness of an intervention/treatment
Number of people that need to receive the intervention before 1 person benefits from it
How to calculate NNT?
1/(Risk difference in terms of proportion)
What two tests could you use to compare a binary variable between two independent groups?
Chi-squared test
Or
Fisher’s exact test
When would you use the Chi-squared test?
The total sample size is at least 40
OR
The sample size is at least 20 and there is an expected value of at least 5 in each cell
When would you use Fisher’s exact test?
When the parametric assumptions for the chi-squared test do not hold
How is the expected value for a cell calculated?
((Row total) X (Column total))/Total sample size
What can proportion be used to infer?
The proportion tells you the risk of one person in a population having a disease
If the odds of having a disease is 1.5, what does this mean about the population?
For every 1.5 people with the disease, 1 person does not have it
