3

The following convention is sometimes encountered in the text (here example from HMM statement, $\{X_n\}_{n \geq 1}$ is a Markov process):

$$X_n \mid (X_{n-1} = x_{n-1}) \sim f(x_n \mid x_{n-1})$$

It is clear, what it means for the "plain" case (Why are probability distributions denoted with a tilde?), but what about the meaning of full formal notation when we have the conditional part $\mid\ (X_{n-1} = x_{n-1})$?

Community
  • 1
Roman Susi
  • 145
  • 1
  • 5

1 Answers1

3

This might be translated to English as:

The distribution of $X_n$, conditional on $X_{n-1}$ taking the value $x_{n-1}$, is described by the probability density function $f(x_n \mid x_{n-1})$.

Matthew Gunn
  • 20,541
  • 1
  • 47
  • 85
  • Thanks, could you please also write it in another notation, as I believe the original is just an abbreviation, or is the $X|Y$ the most used one? (If my guess is right, it should be something with p()) – Roman Susi Dec 17 '15 at 09:03
  • $X\mid Y$ is standard. For example, $P( X=x \mid Y =y)$, often written $P(X\mid Y)$ is the probability that $X$ takes the value $x$ given $Y$ takes the value $y$. $E[X \mid Y=y]$ is the expectation of $X$ conditional on $Y$ takes value $y$. $X \mid Y $ in general refers to $X$ conditional on $Y$. – Matthew Gunn Dec 17 '15 at 09:07
  • Ok. Those others are of course very familiar, but the version without any "functors" around was confusing. – Roman Susi Dec 17 '15 at 09:11
  • @RomanSusi No worries. There's a lot of notation around the use of probability. We've all had our confused "uhhh, so what is this?" moments! – Matthew Gunn Dec 17 '15 at 09:14
  • Even better would be to translate it to English as: "The distribution of $X_n$, conditional on $X_{n-1}$ taking the value $x_{n-1}$, is described by the probability density function $f(\ \mid x_{n-1})$." – Did Dec 27 '15 at 18:05