Scribd
Upload
7 views30 pages

Simple Models

The document discusses simple models in computational psychology, focusing on estimating a single parameter to minimize aggregate errors. It highlights the importance of different types of aggregate errors and their influence on summary measures, such as measures of central tendency and variability. Key points include the choice of error method affecting the best estimate and the use of typical errors like standard deviation for data representation.

Uploaded by

vadodec786
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
7 views30 pages

Simple Models

The document discusses simple models in computational psychology, focusing on estimating a single parameter to minimize aggregate errors. It highlights the importance of different types of aggregate errors and their influence on summary measures, such as measures of central tendency and variability. Key points include the choice of error method affecting the best estimate and the use of typical errors like standard deviation for data representation.

Uploaded by

vadodec786
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

HSO 402

Computational Psychology
Simple Models
Simple Models
• Models with one or no parameters

β0 is the parameter (it is to be estimated)


Determining the parameter
• For a simple model only one true parameter has to be estimated

• For this we need the estimate to minimize the errors


Determining the parameter
• For a simple model only one true parameter has to be estimated

• For this we need the estimate to minimize the errors

• We are not interested in each individual error but in aggregate


of all errors in the data

• Type of aggregate error will influence the type of summary


measure (the model)
Topics to be covered...
1. Types of Aggregate Errors

2. Summary Measures

3. Typical Errors
Conceptual Alignment
• Estimate parameter in a simple model is what is known as
• Measures of Location
• Measures of Central Tendency

• Typical Errors are


• Measures of Variability
• Measures of Spread
Conceptual Example
STUDENT Number of Books Read in
the Vacation

A 1
B 3
C 5
D 9
E 14

How many books does a student read over a vacation ?


Conceptual Example
STUDENT Number of Books Read in
the Vacation

A 1
B 3
C 5
D 9
E 14

A simple model will make the same prediction for all students
i.e. all students read same number of books
But there is random variation in the data

The same prediction (constant estimate) is the parameter


STUDENT Number of Books Read Error Error
in the Vacation (under estimate 5) (under estimate 7)

A 1 4 6
B 3 2 4
C 5 0 2
D 9 4 2
E 14 9 7
Sum of Absolute 19 21
Errors (SAE)
Different Error - Squares rather than absolute values
STUDENT Number of Books Read Error Error
in the Vacation (under estimate 5) (under estimate 7)

A 1 16 36
B 3 4 16
C 5 0 4
D 9 16 4
E 14 81 49
Sum of Squared Errors 117 109
(SSE)
Key points
• Best estimate (b0) for the simple (one parameter) model is the
constant prediction

• Choice of method for expressing the error influences the best


estimate:
• Absolute errors – Median is the best estimate
• Squared errors – Mean is the best estimate
We need to select the estimate such that:

• It is representative of the data (center of the data)


• Minimizes the error
Count of Errors (CE)

Sum of Errors (SE)


Sum of Absolute Errors (SAE)

Sum of Squared Errors (SSE)


Typical Errors
• Data is usually described with a measure of central tendency
and a ‘typical’ error rather than aggregate error

• Aggregate error is too sensitive to size of the data - becoming


unsuitable

• Common Typical Errors


• Modal Error (based on CE)
• Mean Absolute Deviation (based on SAE)
Standard Deviation
• The most commonly occurring typical error

• Based on Sum of Squared Errors (SSE)


Standard Deviation
• Mean Squared Error
n = number of
observations

p = number of
parameters

n - p is the number of additional parameters that could


have been used to reduce the error

You might also like