# How is the notation $X\sim N(\mu,\sigma^2)$ read?

How is the notation $X\sim N(\mu,\sigma^2)$ read? Is it $X$ follows a normal distribution? Or $X$ is a normal distribution? Or perhaps $X$ is approximately normal..

What if there are several variables that follow (or whatever the words is) the same distribution? How is it written?

• $X\sim N(\mu,\sigma)$ should be $X\sim N(\mu,\sigma^2)$ Commented Jul 16, 2015 at 16:53
• @mandata that (unfortunately) depends on who you ask. Many authors use $\sigma$ in both definition and notation.
– KOE
Commented Jul 16, 2015 at 16:57
• Common notation is that "$\sim$" means distributed as, "$\dot \sim$" (note the dot) means approximately distributed as. Commented Jul 16, 2015 at 17:01
• @Student001, because it's prettier and because it's standard (en.wikipedia.org/wiki/Normal_distribution#Notation). $N$ looks like any other letter. $\mathcal N$ is immediately recognized as a normal distribution. Commented Jul 17, 2015 at 16:01
• @amoeba well, my point is that the first part is just your personal preference, and the part about it being standard is not true. The capital N is a common notation.
– KOE
Commented Jul 17, 2015 at 16:06

The variable X is distributed according to the Normal distribution with mean vector $$\mu$$ and standard deviation $$\sigma$$.

• Why vector $\mu$?
– not
Commented Jul 16, 2015 at 17:05
• Because the normal distribution can be multivariate. It can be single value, it can also be generalized to $n$ dimensions. Commented Jul 16, 2015 at 17:06
• Why is the $\sigma$ only a scalar?
– not
Commented Jul 16, 2015 at 17:06
• You are right, the $\sigma$ is not scalar in general for multivariate case. You are speaking then about the covariance matrix $\Sigma$ Commented Jul 16, 2015 at 17:08
• standard deviation. Commented Jul 17, 2015 at 18:55

As regards the use of symbols $\sim$ ("follows", "is distributed according to "), and $\approx$ ("equals approximately"), see this answer. This is how the symbols are used at least in Statistics/Econometrics.

As regards the notational conventions for a distribution, the normal is a borderline case: we usually write the defining parameters of a distribution alongside its symbol, the parameters that will permit one to write correctly its Cumulative distribution function and its probability density/mass function. We do not note down the moments, which usually are a function of, but not equal to, these parameters.

So for a Uniform that ranges in $[a,b]$ we write $U(a,b)$. The mean of the distribution is $(a+b)/2$ while the variance is $(b-a)^2/12$. For a Gamma (shape-scale parametrization), we write $G(k,\theta)$. The mean is $k\theta$ and the variance $k\theta^2$. Etc.

In the case of the normal distribution, the parameter $\mu$ happens to also be the mean of the distribution, while the parameter $\sigma$ happens to be the square root of the variance. It is my (possibly mistaken) impression that in Engineering circles one sees more often $N(\mu, \sigma)$ (which conforms with the general notational rule), while in Econometrics circles almost always one sees $N(\mu, \sigma^2)$ (which falls to the temptation of providing the moments, by treating $\sigma^2$ as the base parameter and not as the square of it).

EDIT: My previous answer failed to answer the actual question. What follows is my attempt at a more to the point response.

How is the notation $$X \sim N(\mu,\sigma^2)$$ read?

Other answers already tell you what the notation means, namely that $$X$$ is a normally distributed random variable with some mean $$\mu$$ and variance $$\sigma^2$$. Dilip's answer also gives a nice account of what other possible interpretations there are when the notation is less clear than $$\sigma^2$$, e.g. for general parameters $$\{a,b\}$$, viz. $$X\sim N(a,b)$$.

Whenever I see this notation in text I tend to read it so that it makes sense grammatically. I would claim that this the sensible way to treat the notation. Thus, the answer to your question is that, knowing what the notation means mathematically, you simply read it in any way that fits the text. Here are a two examples:

(1) Let $$X \sim N(a,b)$$...

(2) Consider three independent random variables, $$X\sim N(0,1), Y\sim N(1,2), Z \sim Exp(\lambda).$$

In (1) I read it as (e.g.) "Let $$X$$ be normally distributed with mean a and variance b...", and in (2) I read it as "... $$X$$ is standard normal...".

Is it X follows a normal distribution?

Yes that works, too. Many people say it this way, although you might want to include the mean and variance characterizing the distribution.

Or X is a normal distribution?

No, that is incorrect. See this old answer of mine for an account of what a distribution is.

Or perhaps X is approximately normal..

No, that is also incorrect. There are other ways to denote this. As pointed out in the comments, $$\overset{\cdot}{\sim}$$ is one of them.

What if there are several variables that follow (or whatever the words is) the same distribution? How is it written?

If they are all independent, one easy way to write this is $$X_i \overset{iid}{\sim} N(\mu,\sigma^2),i=1,2,\dots n$$, given that you have $$n$$ variables (iid stands for independent and identically distributed). If they are not independent, you can say that $$X_i, i=1,2,\dots,n$$ are possibly dependent, but (marginally) identically distributed as $$N(\mu,\sigma^2)$$. Or you may have to instead declare their joint distribution -- that depends on what purpose you have for considering the random variables.

If they are jointly normal, it's easy to write that $$\mathbf X :=(X_1,\dots,X_n)'\sim N(\mu, \Sigma)$$ to fully characterize their joint distribution using some mean vector $$\mu$$ and covariance matrix $$\Sigma$$.

In general, you may define any multivariate distribution function $$F$$ and then write that $$\mathbf X \sim F$$.

• Isn't it nice that regardless of the convention used, $\mathcal N(0,1)$ is always the standard normal random variable? Commented Jul 18, 2015 at 22:10
• @DilipSarwate, indeed! Makes the name "standard" very suitable, too.
– KOE
Commented Jul 18, 2015 at 22:13

The difficulty is not in knowing what $\mathcal N(\mu,\sigma^2)$ means. Even $\mathcal N(3,5^2)$ is reasonably unambiguous to most peaople as meaning a normal random variable with mean $3$ and variance $5^2$ or variance $25$ (purists should believe that the standard deviation is a more fundamental parameter than the variance should free to say "standard deviation $5$" instead). However, what is meant by $\mathcal N(a,b)$, e.g. $\mathcal N(3,25)$ is subject to at least three different conventions with respect to the variance or standard deviation. All three conventions agree that the $\mathbf 3$ is the mean $\mu_X$ of $X$ but the $\mathbf 25$ has different meanings to different people.

• $X \sim \mathcal N(\star,25)$ means that the standard deviation of $X$ is $25$.

• $X \sim \mathcal N(\star,25)$ means that the variance of $X$ is $25$.

• $X \sim \mathcal N(\star,25)$ means that the variance of $X$ is $\dfrac{1}{25}$.

• who besides you, ever had the interpretation that the 2nd parameter of a Normal is the inverse of the variance? This is the first time I recall seeing such a thing. Commented Jul 19, 2015 at 14:23
• @MarkL.Stone Please don't cast aspersions on my veracity. If you had bothered to follow the link that I have included in my answer and read the comments, you would have seen that Moderator whuber said "Others, especially in a Bayesian context, even parameterize Normals by their precision, as in $N(\mu,1/\sigma^2)$." and Moderator cardinal said "there are also the natural parameters of the normal, which probably look quite unnatural to most." These "natural parameters" arise when the normal distribution is defined as a member of the exponential family of distributions. Commented Jul 19, 2015 at 15:47
• I wasn't trying to cast aspersions on your veracity. I looked at the thread and saw your answer, but missed whuber's comment. I guess I'm not a Bayesian. Commented Jul 19, 2015 at 16:27

$X$ is a random variable "$X$";

$\sim$ is read "is distributed as";

$N$ is read "Normal";

$\mu$ is read "with mean $\mu$" (the convention is that the first entry after the open parenthesis is the mean, and the second is the variance or standard deviation, depending on notation -- see below); and

$\sigma^2$ is read "with variance $\sigma^2$ (or standard deviation $\sigma^2$, depending on the usage of the author/user. In this case, I'm guessing it's with variance $\sigma^2$.

Putting it all together, you have a random variable $X$ which is distributed as Normal with a mean "mu" ($\mu$) and variance "sigma squared" ($\sigma^2$).

You can also say $X$ follows a normal. . .

If several variables follow the same distribution, you can represent this several ways, but you might want to index the variables from $i=1$ to $n$. Then you could write, $X_i\sim N(\mu, \sigma^2)$, for $i=1$ to $n$.

$X$ is Normally distributed with mean $\mu$ and standard deviation $\sigma$. The tilde does not mean approximation, as it is not related to an equals sign, though implies it in a way since X is never definitively known.

• Why not? There are populations that are entirely known.
– not
Commented Jul 16, 2015 at 17:04
• $X$ represents a variable, not a set of values. Commented Jul 16, 2015 at 17:40
• X is indeed a random variable and x might be one of its values. But that means there's no approximation: everything there is to (definitively) know about X is stated in the expression we're discussing. Commented Jul 17, 2015 at 18:58
• For the record, $\sim$ is a tilde. Tilda is a brand of basmati rice :-) Commented Jul 17, 2015 at 19:03