Statistics for Machine Learning A-Z
Last Updated on June 12, 2022 by Editorial Team
Author(s): Gencay I.
Originally published on Towards AI the World’s Leading AI and Technology News and Media Company. If you are building an AI-related product or service, we invite you to consider becoming an AI sponsor. At Towards AI, we help scale AI and technology startups. Let us help you unleash your technology to the masses.
∘ Numerical Variable
∘ Categorical Variable
∘ Continuous Variable
∘ Discrete Variables
∘ Dependent Variable:
∘ Independent Variable
∘ Observational Studies
∘ Experimental Studies
∘ Simple Random Sample
∘ Stratified Sample
∘ Placebo Effect
∘ Dot plot
∘ Left skewed
∘ Right skewed
∘ Standard deviation
∘ Null Hypothesis
∘ Alternative Hypothesis
∘ Law of Large Numbers
∘ Mutually Exclusive ( Disjoint)
∘ Probability Trees
∘ Normal Distribution
∘ Binomial Distribution
∘ Bernoulli Distribution
∘ PDF (Probability Density Function)
∘ Z Score
∘ Sampling Variability
∘ Central Limit Theorem
∘ Confidence Interval
∘ Significance Level
∘ Statistical Inference
∘ Type 1 Error
∘ Type 2 Error
∘ T Distribution
∘ Degrees of Freedom
Programming, Statistics, Calculus.
These are 3 things that you should be familiar with if you would like to be involved in Machine Learning.
Image By Author
While there are too many courses that existed in the market, I love creating that kind of article to remind myself of these terms.
That helps me refresh my memories and make repetition.
Repetition is the mother of learning, the father of action, which makes it the architect of accomplishment.” Zig Ziglar
Let’s dive into these terms from beginning to intermediate.
The value that contains an integer.
Contains categories instead of numbers, such as human body shapes such as; skinny, fat, or muscular.
1, 2, 3,4, 5,6, 7 … Take a number of values in a given range.
1, 5, 8, 11, 35 . The specific set of values.
The two variables, when one changes if the other will change will be dependent variables.
If others won't change, independent.
The methods won't be specified by researchers. For example, when they asked you about the method of losing weight, they do not offer you the method such as diet or sport, you could say whatever you like.
Now there are limited options, choose one, diet or sport.
Simple Random Sample
You could choose anything.
Split populations into the clusters, then randomly sample from each cluster.
You will use fake care.
Could we draw a conclusion as a result of our data on the population?
It provides a useful view of data density.
If your sample size is small and you want to view individual data points.
It is good to see statistical values such as IQR and median.
Interquartile range,Range of thee middle 50 %, Q3-Q1.
The tail will be on the right side and the density, mean< median.
The tail will be on the right, mean > median.
The mean and median are close together.
is the number that exists in the middle.
1,5,7 , median : 5
Sum and divide by the number of integers,
The average squared deviation of the mean.
n: number of sample
Variance = (Number1-mean)**2 + (number2 -mean) ** 2 …. + (number n-mean) **2 / (n-1)
The square root of the variance.
Standard deviation : (Variance) ** 1/2
Most frequent number.
1,4,4, 7 , mode : 4
Nothing going on, everything should be the same.
Something going on, something should be changed.
The possibility of your null hypothesis is true.
If the p-value < 0.05, you would reject the null hypothesis, and accept the alternative:
If the p-value >0.05, you would reject the alternative hypothesis.
Law of Large Numbers
As the sample size increases, the mean would be closer to the population means.
Mutually Exclusive ( Disjoint)
Cases that can not happen at identical times.
Cases that could happen at an identical time.
The trees of continuous possibilities.
It is a probability distribution, that shape is symmetric around the mean.
Probability of success or failure ( 2 possible outcomes, like heads or tail. )
It is a discrete distribution, and we have still 2 possible outcomes, 0 or 1.
PDF (Probability Density Function)
The function provides the possibility of the value of a random variable will be in the predefined range.
(observation-mean) / SD
Z score of the mean is zero.
Sometimes images are stronger than words for describing.
It is impossible to gather data from the whole population, we will gather information from different samples and sampling them together via this.
Central Limit Theorem
Describes shapes centers and spreads of sampling distributions when certain conditions are matched. (nearly normal population)(center-mean)(spread-SE)
A reasonable range of values for the population parameter is called a confidence interval.
%90 confident of the “ Estimate ±1.65∗SE(Estimate)”
%95 confident that the “ Estimate ±1.96∗SE(Estimate) = (x−1.96σ/√n, x+1.96σ/√n)
% 99 confident that the “ “Estimate ±2.58∗SE(Estimate)”
The probability of rejecting Ho when you shouldn't do.
the likelihood of rejecting the null hypothesis when it should be rejected.
How close is your estimation to the true value?
It is a quality measurement, how close will your two measurements be to each other.
Finding a conclusion from your data by using statistics.
Type 1 Error
It occurs when rejecting a null hypothesis but it should not have to be rejected.
Type 2 Error
It occurs when the null hypotheses should be rejected but you don't.
Comparing means while standard deviation is unknown.
Degrees of Freedom
Determines thickness of tails.
As the degrees of freedom increase, the shape of the t-distribution approaches the normal distribution.
There are too many terms that existed which I maybe write another article in near future.
That depends both on the statistics of that article and my path.
If you want a follow-up article to that one, do not forget “thumbs up” and follow me.
“Machine learning is the last invention that humanity will ever need to make.” Nick Bostrom
Join thousands of data leaders on the AI newsletter. It’s free, we don’t spam, and we never share your email address. Keep up to date with the latest work in AI. From research to projects and ideas. If you are building an AI startup, an AI-related product, or a service, we invite you to consider becoming a sponsor.
Published via Towards AI