Order Scoring and Structure Creation

[efrain.gonzalez0]

Good morning,

Firstly, there has been a change to the way that priors must be set. Recall that previously one had to specify the priors in the following way:

Code: Select all

1,2(tab).9(tab)0,4(tab).1

Now the format for the above priors would be set in the following way:

Code: Select all

1,2(tab)0,4(tab).9

Both of the above statements mean the same thing but the latter is the format that you will need to use in your prior file.
Answers to questions for setting up your prior file:

1. What do the above two statements mean? Both statements are saying that 1 and 2 come before 0 and 4 in the order with probability .9.
2. Can this be written in another way? Yes, an equivalent statement would be the following:
   

   Code: Select all

   0,4(tab)1,2(tab).1

   
   The above statement means that 0 and 4 come before 1 and 2 in the order with probability .9.
3. Can I use 1 or 0 as my probability? No, currently this may result in a divide by zero error so I would not use 1 or zero until I have updated the current program.
4. If I use a probability equal to .9 then will the probability of the reverse happening be .1? Yes, in general if you use a probability p of the first set coming before the second set then the probability of the second set coming before the first set will be 1-p.
5. Do I need to specify a probability for all variables in my data set? No, in fact this will not be possible for most data sets because we will not have sufficient information to specify the priors with such detail.
6. What happens if I do not specify a variable within the prior file? If a variable is not specified within the prior file then the probability of that variable being in front of another variable in the order is considered to be .5. Meaning that every time that variable is being considered we are tossing a coin.

Below I have attached some results for the MCMC tests that I have been running. Recently, I noticed another problem with the cutdeck feature that I have included and so that feature will not be available just yet.

Respectfully,

Efrain Gonzalez

[efrain.gonzalez0]

Good morning,

There is a new version of the MCMC Order Scoring code that has been posted on smlg.fiu.edu/gitlab under the Cpp_Programming/Order Score.
What is different about this new version?:

1. Now you can use REST as a way of defining a set. With this edition the following two prior sets are the same for a data set with four variables:
   

   Code: Select all

   Ex1:
0,1(tab)2,3(tab).9
Ex2:
0,1(tab)REST(tab).9


   
   The idea is that this is beneficial when one has a large amount of variables.
   One issue is that if you say REST REST an error occurs. However in the future any line in the prior file that has REST as the first set will be ignored. The following are examples of the type of inputs that will be ignored.
   

   Code: Select all

   Ex3:
REST(tab)2,3(tab).9
Ex4:
REST(tab)REST(tab).9


   
   2. probability equals 1 not handled yet
   3. use probability equal to .5 when an error occurs
2. You are now allowed to set priors equal to 1 or 0. The following are examples of this type of input:
   

   Code: Select all

   Ex5:
0,1(tab)REST(tab)1


   
   This means that 0 and 1 should always be considered to be at the beginning of the order.
3. The new version also allows you to remove variables from the list found in REST. This functionality is shown below:
   

   Code: Select all

   Ex6: Without the functionality
0,1(tab)2(tab).9
2(tab)3(tab).9

Ex7: With the functionality
0,1(tab)REST(tab).9
2(tab)REST,-0,-1(tab).9