Model thinking lecture note (3)

Rational actor model

We assume that

(1)actor has his own objective

(2)actor will optimize his objective

Then we can predict what other people would do, Cause they're rational. Also we can also use this assumption to know what we are going to do next.

Rule based model

we can divide the model into 4 parts

(1) Decision fixed rule

(2)Decision adaptive rule

(3)Game fixed rule

(4)Game adaptive rule

where decision means we make our choice depending only on ourselves while game means we make our choice depending on others

Categorical model

This model can help us filter the data into several categories.

(1) we can compute the data set's variance V1

(2)From seeing how much the data getting away from mean value , we can make the data set into several categories. Then we can compute their variance V2 V3 ... ...

(3)we can compute the variation explained 1 - (V2 + V3 +... ...)/V1

Linear model

Sometimes the linear model (one input one output linear model means just a line) can be better than expert

We can use linear regression method to get a straight line.

How can we see the output of regression

Standard error: the sigma mentioned in the central theorem. It can show how many data points deviated from  the mean.

Observation: how many data sets in the model.

We can make our decision based on the sign and the magnitude of the coefficient of the model. The sign is minus that means one feature declining the output, and the magnitude shows how strong this feature is. 

P_value: what's the probability of this value being minus.