Probability distribution of a continuous random variable
Continuous random variables, unlike discrete random variables, can take any value in a real interval. Thus the range of a continuous random variables is infinite and uncountable.
Such a density of values makes impossible to compute the probability for each one of them, and therefore, it’s not possible to define a probabilistic model trough a probability function like with discrete random variables.
Besides, usually the measurement of continuous random variable is limited by the precision of the measuring instrument. For instance, when somebody says that is 1.68 meters tall, his or her true height is no exactly 1.68 meters, because the precision of the measuring instrument is only cm (two decimal places). This means that the true height of that person is between 1.675 y 1.685 meters.
Hence, for continuous variables, it makes no sense to calculate the probability of an isolated value, and we will calculate probabilities for intervals.
Probability density function
To model the probability distribution of a continuous random variable we use a probability density function.
Definition  Probability density function. The probability density function of a continuous random variable $X$ is a function $f(x)$ that meets the following conditions:
 It is nonnegative: $f(x)\geq 0$ $\forall x\in \mathbb{R}$,
 The area bounded by the curve of the density function and the xaxis is equal to 1, that is,
 The probability that $X$ assumes a value between $a$ and $b$ is equal to the area bounded by the density function and the xaxis from $a$ to $b$, that is,
The probability density function measures the relative likelihood of every value, but watch out!, $f(x)$ is not the probability of $x$, cause $P(X=x)=0$ for every $x$ value by definition.
Distribution function
The same way that for discrete random variables, for continuous random variables it makes sense to calculate cumulative probabilities.
Definition  Distribution function. The distribution function of a continuous random variable $X$ is a function $F(x)$ that maps every value $a$ to the probability that $X$ takes on a value less than or equal to $a$, that is,
Probabilities as areas
To calculate probabilities with a continuous random variable we measure the area bounded by the probability density function and the xaxis in an interval.
This area can be calculated integrating the density function or substracting the distribution function that is easier,
Example. Given the following function
let’s check that is a density function.
As this function is clearly nonnegative, we have to check that total area bounded by the curve and the xaxis is 1.
Now, let’s calculate the probability of $X$ having a value between 0 and 2.
Population statistics
The calculation of the population statistics is similar to the case of discrete variables, but using the density function instead of the probability function, and extending the discrete sum to the integral.
The most important are:

Mean:

Variance:

Standard deviation:
Example. Let $X$ be a variable with the following probability density function
The mean is
and the variance is
Continuous probability distribution models
According to the type of experiment where the random variable is measured, there are different probability distributions models. The most common are
 Continuous uniform.
 Normal.
 Student’s T.
 Chisquare.
 FisherSnedecor’s F.
Continuous uniform distribution
When all the values of a random variable $X$ have equal probability, the probability distribution of $X$ is uniform.
Definition – Continuous uniform distribution $U(a,b)$. A continuous random variable $X$ follows a probability distribution model uniform of parameters $a$ and $b$, noted $X\sim U(a,b)$, if its range is $\mbox{Ran}(X) = [a,b]$ and its density function is
Observe that $a$ and $b$ are the minimum and the maximum of the range respectively, and that the density function is constant.
The mean and the variance are and .
Example. The generation of a random number between 0 and 1 is follows a continuous uniform distribution $U(0,1)$.
As the density function is constant, the distribution function has a linear growth.
Example. A bus has a frequency of 15 minutes. Assuming that a person can arrive to the bus station in any time, what is the probability of waiting for the bus between 5 and 10 minutes?
In this case, the variable $X$ that measures the waiting time follows a continuous uniform distribution $U(0,15)$ as any waiting time between 0 and 15 is equally likely.
Then, the probability of waiting between 5 and 10 minutes is
And the expected waiting (the mean) time is $\mu=\frac{0+15}{2}=7.5$ minutes.
Normal distribution
The normal distribution model is, without a doubt, the most important continuous distribution model as it is the most common in Nature.
Definition  Normal distribution $N(\mu,\sigma)$. A continuous random variable $X$ follows a probability distribution model normal of parameters $\mu$ and $\sigma$, noted $X\sim N(\mu,\sigma)$, if its range is $\mbox{Ran}(X) = (\infty,\infty)$ and its density function is
The two parameters $\mu$ and $\sigma$ are the mean and the standard deviation of the population respectively.
The plot of the probability density function of a normal distribution $N(\mu,\sigma)$ is bell shaped and it is known as a Gauss bell.
The bell shape depends on the mean $\mu$ and the standard deviation $\sigma$,
 The mean $\mu$ sets the center of the bell.
 The standard deviation sets $\sigma$ the width of the bell.
The plot of the distribution function of a normal distribution is S shaped.
Normal distribution properties
 It is symmetric with respect to the mean, and therefore, the coefficient of skewness is zero, $g_1=0$.
 It is mesokurtic, as the density function is bell shaped, and so, the coefficient of kurtosis is zero, $g_2=0$.
 The mean, median and mode are the same
 It asymptotically approaches 0 when $x$ tends to $\pm \infty$.
 $P(\mu\sigma \leq X \leq \mu+\sigma) = 0.68$
 $P(\mu2\sigma \leq X \leq \mu+2\sigma) = 0.95$
 $P(\mu3\sigma \leq X \leq \mu+3\sigma) = 0.99$
Example. It is known that the cholesterol level in females of age between 40 and 50 follows a normal distribution with mean 210 mg/dl and standard deviation 20 mg/dl.
According to the Gauss bell properties, this means that
 The 68% of females have a cholesterol level between $210\pm 20$ mg/dl, i.e., between 190 and 230 mg/dl.
 The 95% of females have a cholesterol level between $210\pm 2\cdot 20$ mg/dl, i.e., between 170 and 250 mg/dl.
 The 99% of females have a cholesterol level between $210\pm 3\cdot 20$ mg/dl, i.e., between 150 and 270 mg/dl.
Example of blood analysis. In blood analysis it is common to use the interval $\mu\pm 2\sigma$ to detect possible pathologies. In the case of cholesterol, this interval is $[170\text{ mg/dl}, 250\text{ mg/dl}]$.
Thus, when a women between 40 and 50 years of age has a cholesterol level out of this interval, it’s common to think about some pathology. However this person could be healthy, although the likelihood of that happening is only 5%.
The central limit theorem
This behavior is common in many physical and biological variables in Nature.
If you think about the distribution of the height, for instance, you can check that most people in the population have a height around the mean, but as the heights move away from the mean, both below and above the mean, there are few and few people with such a heights.
The explanation for this behavior is the , that we will see in the next chapter; it states that a continuous random variable whose values depends on a huge number of independent factors adding their effects, always follows a normal distribution.
The standard normal distribution $N(0,1)$
The most important normal distribution has mean zero, $\mu=0$, and standard deviation one, $\sigma=1$. It is known as Standard normal distribution and usually represented as $Z\sim N(0,1)$.
Calculation of probabilities with the normal distribution
To avoid integrating the normal density function to compute probabilities it’s common to use the distribution function, that is given in a tabular format like the one below. For instance, to calculate $P(Z\leq 0.52)$
0.00  0.01  0.02  …  

0.0  0.5000  0.5040  0.5080  … 
0.1  0.5398  0.5438  0.5478  … 
0.2  0.5793  0.5832  0.5871  … 
0.3  0.6179  0.6217  0.6255  … 
0.4  0.6554  0.6591  0.6628  … 
0.5  0.6915  0.6950  0.6985  … 
⋮  ⋮  ⋮  ⋮  ⋮ 
To compute cumulative probabilities to the right of a value, we can apply the rule for the complement event. For instance,
Standardization
We have seen how to use the table of the standard normal distribution function to compute probabilities, but, what to do when the normal distribution is not the standard one?
In that case we can use standardization to transform any normal distribution in the standard normal distribution.
Theorem  Standardization. If $X$ is a continuous random variables that follow a Normal probability distribution model with mean $\mu$ and standard deviation $\sigma$, $X\sim N(\mu,\sigma)$, then the variable that result of subtracting $\mu$ to $X$ and dividing by $\sigma$, follows a Standard Normal probability distribution,
Thus, to compute probabilities with a nonstandard normal distribution first we have to standardize the variable before using the table of the standard normal distribution function.
Example. Assume that the grade of an exam $X$ follows a normal probability distribution model $N(\mu=6,\sigma=1.5)$. What percentage of students didn’t pass the exam?
As $X$ follows a nonstandard normal distribution model, we have to apply standardization first, $Z=\displaystyle \frac{X\mu}{\sigma} = \frac{X6}{1.5}$,
Then we can use the table of the standard normal distribution function,
Therefore, $25.14\%$ of students didn’t pass the exam.
Chisquare distribution
Definition  Chisquare distribution $\chi^2(n)$. Given $n$ independent random variables $Z_1,\ldots,Z_n$, all of them following a standard normal probability distribution, then the variable
follows a chisquare probability distribution with $n$ degrees of freedom.
Its range is $\mathbb{R}^+$ and its mean and variance are $\mu = n$ and $\sigma^2 = 2n.$.
Example. Below are plotted the density functions of some chisquare distribution models.
Chisquare distribution properties
 The range is nonnegative.
 If $X\sim \chi^2(n)$ and $Y\sim \chi^2(m)$, then .
 It asymptotically approaches to a normal distribution as the degrees of freedom increase.
As we will see in the next chapter, the chisquare distribution plays an important role in the estimation of the population variance and in the study of relations between qualitative variables.
Student’s t distribution
Student’s t distribution $T(n)$. Given a variable $Z$ following a standard normal distribution model, $Z\sim N(0,1)$, and a variable $X$ following a chisquare distribution model with $n$ degrees of freedom, $X\sim \chi^2(n)$, independent, the variable
follows a Student’s t probability distribution model with $n$ degrees of freedom.
Its range is $\mathbb{R}$ and its mean and variance are and if $n>2$.
Example. Below are plotted the density functions of some student’s t distribution models.
Student’s t distribution properties
 The mean, the median and the mode are the same, $\mu=Me=Mo$.
 It is symmetric, $g_1=0$.
 It asymptotically approaches to the standard normal distribution as the degrees of freedom increase. In practice for $n\geq 30$ both distributions are approximately the same.
As we will see in the next chapter, the Student’s t distribution plays an important role in the estimation of the population mean.
FisherSnedecor’s F distribution
FisherSnedecor’s F distribution $F(m,n)$. Given two independent variables $X$ and $Y$ both following a chisquare probability distribution model with $m$ an $n$ degrees of freedom respectively, $X\sim \chi^2(m)$ and $Y\sim \chi^2(n)$, then the variable
follows a FisherSnedecor’s F probability distribution model with $m$ and $n$ degrees of freedom.
Its range is $\mathbb{R}^+$ and its mean and variance are and if $n>4$.
Example. Below are plotted the density functions of some FisherSnedecor’s F distribution models.
FisherSnedecor’s F distribution properties
 The range is nonnegative.
 It satisfies Thus, if we name $f(m,n)_p$ the value that satisfies $P(F(m,n)\leq f(m,n)_p)=p$, then which is helpful in order to compute probabilities from the table of the distribution function.
As we will see in the next chapter, the FisherSnedecor’s F distribution plays an important role in the comparison of population variances and in the analysis of variance test (ANOVA).