Scribd
Upload
20 views23 pages

Introduction To Probabilities, Bayesian and Frequentist Statistics

Uploaded by

asila.edu2021
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
20 views23 pages

Introduction To Probabilities, Bayesian and Frequentist Statistics

Uploaded by

asila.edu2021
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 23

Introduction to

Probabilities, Bayesian and


Frequentist Statistics

Moutoshi Pal, Lica Iwaki & Carolin Lingl


Research Methods in Clinical Psychology
October 30, 2019
Outline of the Presentation

1. Quick summary of the paper

2. The Frequentist approach

3. Misconceptions of the p-value

4. The Bayesian approach

5. Benefits of Bayesian Statistics

6. Comparison of the two approaches


1. Quick summary of the paper

● replication crisis
● issues within Frequentist statistics that may have led to the replication crisis
● the alternative—Bayesian statistics
● rather than mere statistical reform, what is needed is a better understanding

➔ each approach provides a different kind of information that is useful for


different aspects of scientific research

(Colling & Szűcs, 2018)


2. Frequentist Statistics
Null Hypothesis Significance Testing (NHST)

"surely the most bone-headedly misguided procedure


ever institutionalized in the rote training of science
students" (Rozeboom, 1997, p. 335).
Issues with Frequentist Statistics

● p-hacking
● Data dredging
● Significance chasing
● Optional stopping (a.k.a. data peaking)
● QPRs (questionable research practices)
2. Frequentist Statistics

p-value: The p value is calculated from the sampling distribution, which


describes what is to be expected over the long run when samples are
tested

➔ Relates to the idea of a large (infinite) number of repeated


experiments (i.e. coin toss)
➔ Given the null, the frequency of finding a significant result
(p < 0.05) is 5% for each test

(Colling & Szűcs, 2018)


2. Frequentist Statistics

Definition of the p-value

A p -value is the probability of the observed data, or more


extreme data, under the assumption that the null hypothesis
is true.
2. Frequentist Statistics

Definition of the p-value

A p -value is the probability of the observed data, or more


extreme data, under the assumption that the null hypothesis
is true.
2. Frequentist Statistics

Definition of the p-value

A p -value is the probability of the observed data, or more


extreme data, under the assumption that the null hypothesis
is true.
2. Frequentist Statistics

Definition of the p-value

A p -value is the probability of the observed data, or more


extreme data, under the assumption that the null hypothesis
is true.
3. Misconceptions of the p-value
Now it’s time
for… a Kahoot
Quiz!!
Two approaches

● Two approaches on evidence and inference:

1. Frequentist approach: probability is a limit of the relative frequency of an


event after many trials (i.e. coin toss)

- The issue: misconceptions led to replication crisis

2. Bayesian approach: an alternative method


The difference

Recall, Frequentist’s definition of p-value:

“The p -value is the probability of the observed data, or more extreme


data, under the assumption that the null hypothesis is true.”

**Emphasis**:
the p-value is not the probability of a theory or hypothesis, but
the probability of the observed data
given the hypothesis

P(D|Ho)
The difference

Bayesian approach: probability is an expression of a degree of belief of an event,


based on prior knowledge (i.e. previous experiments) or personal beliefs

The p- value is the probability of the hypothesis given the data

P(Ho|D)
P(D|Ho) ≠ P(Ho|D)
Frequentist Bayesian
The probability of the observed data/result given that some hypothesis is true is NOT equivalent to
the probability that a hypothesis is true given that some data/result has been observed.
Bayesian Statistics

● As opposed to Frequentist Statistics (which is based on


an infinite number of repeatable events), Bayesian
Statistics relates with subjective belief

● Thus, Bayesian Statistics uses the idea of updating beliefs


with new information when testing a hypothesis

Historical background:

● In the past, Bayesian statistics was often overlooked by social scientists due to its complexity
● More recently, the development of powerful statistical software tools allows for the estimation of
complex models using the Bayesian perspective
Bayesian Statistics
● Uses the idea of updating beliefs with new information when testing a hypothesis

● Start with a belief about how something works (i.e. “eating sushi is dangerous”)

Prior beliefs x Bayes’ Factor = Posterior belief


(= updated, new belief)

Bayes’ Factor: represents the amount of information


that we’ve learned about our hypotheses form the data

● Criticism: prior belief can be different from person to person


○ Includes subjective “beliefs” in calculation (subjective, not objective)
Bayesian Statistics
● Prior Distribution
● Likelihood gives the function of
a parameter given the data
● Data
Bayesian Curves

Uninformative
Prior belief

𝛉 = 0.5
Comparison of Frequentist Statistics & Bayesian

Frequentist Bayesian
1. Parameters are 1. Parameters are
fixed but unknown random and
and data are data are fixed
random 2. Probability is a
2. Probability is a degree of
measure of certainty about
frequency of values
repeated events
Comparison of Frequentist Statistics & Bayesian

Frequentist Bayesian

Confidence Interval: Credible Interval:


“there is a 95% probability “given our observed
that when I compute a data (posterior
confidence interval from distribution), there is
data of this sort, the true a 95% probability that
value of θ will fall within it”. the true value of θ
falls within the
credible region”
Benefits of Bayesian Statistics

1. They are dependent only on the


observed data (and not the data that
might have been collected)
Subjective Priors
Different Priors will lead
2. They are immune to the unknown to different posteriors &
intentions of the researcher conclusions

3. They provide a measure of the strength


of evidence that takes into account both
the null and the alternative.
References
● Colling, L. J., & Szűcs, D. (2018). Statistical Inference and the Replication Crisis. Review of
Philosophy and Psychology, 1-27.

● Rozeboom, W. W. (1997). Good science is abductive, not hypothetico-deductive. What if there were
no significance tests, 335-391.

● Wasserstein, R. L., & Lazar, N. A. (2016). The ASA’s statement on p-values: context, process, and
purpose. The American Statistician, 70(2), 129-133.
T.H.E E.N.D.

You might also like