Ernesto Palafox Malfavón

Probability & Statistics

Theme 6 Sampling Distribution

Random Sample:

A sample is a collection of data obtained by performing repeated tests in an experiment, in order to

achieve representative evidence about the population related with the X characteristic.

A way to obtain a good sample is when the process of sampling provides for each object of the
population, an equal and independent opportunity of being included on the sample. If the population
consists on N objects and we select a simple of size n, the process of sampling should ensure that
every sample of size n had the same probability of being selected. This procedure leads to what is
commonly known as a simple random sample.

There are some situations that arise frequently on the sampling. The first situation presents itself in
several experiments that include random involuctan( fenómenos aleatorios), in Engineering and in
physic sciences. Therefore, the population of our interest, do not refers to quantifiable objects, more
precisely, the population is an infinite number of possible results, for some measurable
characteristics. Usually, it is a physic measure as in the concentration of a pollutant,

If X being a measurable caracteristic . and f ( x ; θ ) the probability density function of the

population distribution. For this kind of population, the next method it is a type of sampling:

In an experiment we assign to X 1 the measureable ‘X’ caracteristic. The experiment is repeated

under the same conditions, providing the X 2 measure. This procedure continues up to obtain n
measures. X 1 , X 2 , … , X n. In this type of sampling, the measures are based on independent events.
Basicly, this is a randoms sample with replacement. Each one of the observations is a random
variable, which distribution’s probability is the exact the same than the distribution’s probability of
the population

A different case presents when we select quantifiable objects, (human beings, animals, electronic
components, etc.) The measurable characteristics of interest can be a specific attribute. We can use
one of the following technics:

The first one, starts having a combination of objects, we take one, and oberve the measurable
characteristic This observation will be named distribution’s probability ‘ X 1 ’, then we took the
object back to the container, after mixing the content, we take a second object.’ X 2 ’ for observation.
The procedure continues until we have the necessary objects: X 1 , X 2 , … , X n .
Once again we mix all the objects on the container, ‘n’ of these are selected one after the other
without replacement, this process shows an observation sample: X 1 , X 2 , … , X n de característica X
.

The second technic is a random sample with replacement. In this technic, the distribution’s
probability of the first answer does not affect the consequent.

A stadistic is any function of a random variable that is observed in the sample, this way we can
asure this function does not contain unknown quantities.

The central limit theorem

If X 1 , X 2 , … , X n ‘n’ random variables, independent and identically distrubuted, with a non

specified probability whit a mean μ and variance σ 2 /n T El promedio muestral
X́ =( X 1 , X 2 , … , X n )/n tiene una distribución con media μ y varianza σ 2 /n that tends to a normal
distribution, when it gets near to∞ .

The random variable ( X́ −μ)/(σ / √ n) has as a limit a standard normal distribution. If the pattern of
the population probability is such as a norma distribution, the normal approximation will work even
with small samples. On the other hand, if the pattern of population has little resemblance with the
normal approximation, it will be useful only for relatively large on ‘n’. In many cases, we can
guaranteion will be real if n>30. Random varaible formula:

X́−μ
Z=
σ /√n

It is used to create inference formulas about , when we know the population variance amount σ 2
The Z variable is N(0,1)when the sample cuando el muestreo is performed on a population that is
normally distributed.

Generation of values for a random variable

There are many methods to generate values that could be following a normal, the basic one is to
reverse the distribution function of normal standard. There are more efficient procedures, like:
transformation of Box-Muller. A faster algorythm coul be: zigurat. A simple approximation of these
methods is the following:

Add 12 uniform deviations (0,1) substract 6 (exact half). The total sum of this 12 values, follow the
Irwin-Hall distribution, we choose 12 values, to give the sum a exactly 1 variance. The random
deviations obtained are found on the next range:(−6,6).

Box-Muller methods explains that if you have two random numbers, U y V uniformly distributed in
(0, 1], then: X and Y are to random standard variables normally distributed:

For example:
 1) The weight in kg. each piece is a random variable X distributed according to:
1 1< x<5
2) f ( x )
{
= 4 ∀ otro x
0
3) What is the mean and variance of the weight of a batch of 100 pieces?

The weight of each piece is a random variable X_i. All these X_i are identically distributed
according to f given and assumed to be independent. Then the total weight of 100 pieces, about 100
≥ 30, is approximately a normal random variable with mean and deviation 100μ_x 10σ_x. From the
distribution we give to the X_i., we estimate that:
μ_x = 3; σ_x = 4 / 3
If we call and the weight of the batch of 100 pieces, then the mean of Y is μ_x μ_y = 100 = 300,
and the variance of Y is σ_y ^ 2 = (10σ_x) ^ 2 = 178
Y = (10sX) 2 = 178.

Distribution of means.

Each sample of size n that can be drawn from a population gives an average. If we consider each of
these measures as values of a random variable we can study its distribution which we call: sampling
distribution of means.

If we have a normal population: N(,) and we take samples of the same size n, the sample
distribution of mean follows a normal distribution too.

If the population does not follow a normal distribution n> 30, using the central .limit theorem also
called: ‘The sampling distribution of means’ is also close to the previous normal

Where Z_i are independent normal random variables with zero mean and variance one.

The fact that the random variable X has this distribution is usually represented as: X ~ χ_k ^ 2.
Where the letter χ is the Greek letter chi:

The density function is:

where Γ is the gamma function

The distribution function is:

We are interested in calculating the chi-square distribution with its fractals. That is, the values ‘x’
such like: P (X   x) is equal to a . the value of an area to  left under the curve of a chi-square
with extra degrees

These are tabulated fractals-

T Distribution

The random variable X has the t-distribution The density function probability is:
The ‘v’parameter is a natural number, and is known by the name of: "degrees of freedom".

We need to calculate the t-distribution are the fractals. That is, the ‘x’ values: P (X   x) is equal to
some .