Das Log-Wahrscheinlichkeit Ratio (LLR) is a statistical measure used to compare the likelihood of two competing hypotheses, often referred to as the Nullhypothese and the alternative hypothesis. In simple terms, it helps determine how much more likely one hypothesis is compared to another based on beobachtete Daten.
Mathematisch wird der Log-Likelihood Ratio als der Logarithmus des Verhältnisses der Wahrscheinlichkeiten der beiden Hypothesen definiert. Es wird ausgedrückt als:
LLR = log(L(H1) / L(H0))
where L(H1) is the likelihood of the data under the alternative hypothesis and L(H0) is the likelihood of the data under the null hypothesis. The logarithm is used to transform the ratio into a more manageable scale, making it easier to interpret and work with, especially when dealing with very small or very large numbers.
Log-Likelihood-Ratios sind in verschiedenen Bereichen besonders nützlich, einschließlich maschinellem Lernen, bioinformatics, and econometrics, as they provide a way to quantify evidence in favor of one hypothesis over another. A positive LLR indicates that the data supports the alternative hypothesis more strongly, while a negative LLR suggests stronger support for the null hypothesis.
Zusätzlich kann der LLR verwendet werden, um p-Werte für Hypothesentests, which helps determine the statistical significance of the observed results. In practical applications, higher absolute values of LLR correspond to stronger evidence against the null hypothesis, guiding researchers and analysts in making informed decisions based on their data.