2a Critical Numbers - Logistic Regression Flashcards
What are the 4 main reasons Regression is carried out for?
1) Model for risk prediction
2) Isolate effect of a single variable
3) Identify multiple independent predictors of outcome and how it affects outcome
4) Covariate adjustment to improve RCT efficiency (Random imbalances)
What equation is the data fit to?
Y= mx + c [Y = a + bx] Y- Outcome M(b) - Main interest and gradient (Coefficient) X - Predictor (Doesn't change) C(a) - Y intercept
What do Y and X signify in regression equation?
Y: Outcome (Dependent variable)
X: Predictor (Ind variable) - Variable using to estimate
When is Logistic regression used?
When the outcome is binary (Y/N, 0/1)
What does the odds ratio mean if its 1 or CI overlaps it at 1
No difference, therefore no significant difference
What happens if the confidence intervals aren’t symmetric?
Alternative difference ongoing, so need to convert to log scale by ln
What does log scale do and how do you interpret it?
Log scale makes it clearer as CI will not be symmetric on OR scale, exponential to make it more interpretable
What else does logistic regression do?
Takes confounders into account
What are the differences in types of logistic regression analysis?
Univariate: Only looks at 1 variable at a time
Multivariate: Allows for joint effects, so is adjusted as takes into effect confounders
Why would CI of odds ratio not be symmetric?
Estimates are calculated on a log scale which would be symmetric
When is the odds ratio referred to as ‘crude’?
When other confounders aren’t taken into account (AKA adjusted)
When would linear regression be performed?
When the data provided is continuous as the outcome
What is the difference between a 9 fold increase and 9% increase?
9-fold increase means 9x
9% means increase by 1.09
