Retrieve intermediate steps applied to one variable from a lost do-file

[Diego Almonacid]

Hi everyone. I really need your help.

Suppose that I have two variables X and X_2. X_2 comes from different operations (sums, divisions, logarithms, etc.) using X itself and some random variables. I saved the dataset, but my computer crashed and I lost the do-file describing how to get X_2 from X. Thus, the only information I have is a dataset with X and X_2

I would like to ask you if there is some way in Stata for retrieving the intermediate steps or, at least, one approximation of how can I get X_2 from X. For example, X_2 = X + log(random_variable) where random_variable has certain characteristics (distribution, mean, sd) and can be simulated using a determined seed.

Thank you.

[Andrew Musau]

Without seeing the contents of these variables, I would say that the complexity increases with the number of operations applied. If you applied only one operation, then it may be possible to determine the operation. With many operations, there may be an unlimited number of operations that can produce the same variable. But you could present a data example of the variables in case you are confident that we are dealing with the former and not the latter.

dataex

[Diego Almonacid]

Thank you Andrew. There were some operations being executed, including operations with random variables. These operations among the random variables and X allowed me to get X_2 from X. X_2 and X are the only information I have.

[Andrew Musau]

You may not be able to get the exact seed, but if you know the random number function, then getting the arguments may still be possible especially if the dataset is large. I would still encourage you to show us these variables and include any information on the functions. Below, I can guess the mean and standard deviation of a normal random variable by summarizing it.

clear
set obs 10000
set seed 07242023
gen X= rnormal(5, 7)
sum X, d

Res.:

. gen X= rnormal(5, 7)


. sum X, d

                              X
-------------------------------------------------------------
      Percentiles      Smallest
 1%    -11.21509      -19.77879
 5%     -6.47545      -18.93595
10%    -3.979005      -17.84546       Obs              10,000
25%      .206667        -16.493       Sum of wgt.      10,000

50%      5.03195                      Mean           5.021514
                        Largest       Std. dev.      7.017917
75%     9.814817       29.83835
90%     13.95592       30.09274       Variance       49.25116
95%     16.62557       30.15853       Skewness       .0133329
99%     21.20141       34.37077       Kurtosis       2.913767