MHB Maximum of the Likelihood estimation

AI Thread Summary
Maximizing the likelihood function allows for the estimation of parameters that best fit a given statistical model to observed data. The maximum likelihood estimator (MLE) provides the most probable values of these parameters, which can be used to describe the underlying population and make predictions. In the example discussed, the MLE for the parameter θ was calculated as 4/5, indicating the most likely distribution of employee start times. This process illustrates how likelihood estimation can be applied, even if the resulting distribution may not always seem realistic. Ultimately, finding the MLE helps in understanding the data and predicting future occurrences.
mathmari
Gold Member
MHB
Messages
4,984
Reaction score
7
Hey! :o

I am looking at the Likelihood function. I have understood how we define that function, but having find the maximum of the Likelihood estimation, what is the meaning of that?

What information do we have, when we have found the $\theta$ so that the Likelihood function L is maximized?

(Wondering)
 
Physics news on Phys.org
mathmari said:
Hey! :o

I am looking at the Likelihood function. I have understood how we define that function, but having find the maximum of the Likelihood estimation, what is the meaning of that?

What information do we have, when we have found the $\theta$ so that the Likelihood function L is maximized?

Hey mathmari! (Smile)

In statistics we typically try to find a probability distribution that describes a population.
It helps us to understand the population, and moreover to make predictions.
Often enough that's a normal distribution with a mean value and a standard deviation.
It's usually fairly straight forward to find an approximation for those particular parameters.

More generally a population is described by the assumption of a distribution combined with the parameters defining it, like a normal distribution characterized by $\mu$ and $\sigma$, but we can also have different distributions, parameters, and it may not be so straight forward to find the parameters.
Maximizing the likelihood function means we find the most likely approximation of parameters given an assumed distribution. (Thinking)
 
I like Serena said:
In statistics we typically try to find a probability distribution that describes a population.
It helps us to understand the population, and moreover to make predictions.
Often enough that's a normal distribution with a mean value combined and a standard deviation.
It's usually fairly straight forward to find and approximation for those particular parameters.

More generally a population is described the assumption of a distribution combined with the parameters defining it, like a normal distribution characterized by $\mu$ and $\sigma$, but we can also have different distributions, parameters, and it may not be so straight forward to find the parameters.
Maximizing the likelihood function means we find the most likely approximation of parameters given an assumed distribution. (Thinking)

Let's consider a specific example.

An employee starts work around 8:00 am The general beginning of duty varies by up to $2$ minutes up or down. We have the following:

[table="width: 700"]
[tr]
[td]$X$ = Beginning of duty (Difference to 8 o'clock in minutes) [/td]
[td]$-2$ [/td]
[td]$-1$ [/td]
[td]$1$ [/td]
[td]$2$[/td]
[/tr]
[tr]
[td]$\mathbb{P}(X=x)$ [/td]
[td]$0.2\, \theta$[/td]
[td]$0.3\, \theta$[/td]
[td]$0.5\, \theta$[/td]
[td]$1-\theta$[/td]
[/tr]
[/table]
For $10$ consecutive working days, the following values have occurred:
\begin{equation*}-1, \ \ \ 2, \ -2, \ -2, \ \ \ 1, \ \ \ 1, \ \ \ 2, \ -1, \ \ \ 1, \ -1\end{equation*}

From these infotmation we get the Likelihood function:
\begin{equation*}L(-1, 2, -2, -2, 1, 1, 2, -1, 1, -1 \mid \theta ) = 0.000135\cdot \theta^8\cdot (1-\theta)^2 \end{equation*}

The maximum Likelihood estimator is $\hat{\theta}=\frac{4}{5}$.

Do we calculate each probabilities $\mathbb{P}(X=x_i)$ with this value of $\theta$ and then we have the predictions what time the duty begins? (Wondering)
 
mathmari said:
Do we calculate each probabilities $\mathbb{P}(X=x_i)$ with this value of $\theta$ and then we have the predictions what time the duty begins? (Wondering)

Yes.
Now we have found the most likely distribution based on an assumed distribution with an unknown parameter.
To be fair, it looks like a somewhat unrealistic distribution. Then again, it's just an example how maximum likelihood works. (Thinking)
 
Back
Top