Simulating data from an unknown distribution given min and max values Let's say I have a couple of studies that have measured the values of a variable in different experimental setups. For example, the first study evaluated the effects of A & B on y and the second study evaluated the effects of C & D on y.  In both studies they authors reported only the min and max of the values ( i.e., no information ragrding mean, sd, or the distribution of y is given). 
I want to make a regression model that accounts for A,B,C,D. Is it a valid procedure to use a triangular distribiution (or any location based distribution) to generate some random data based on the given min and max values and then make a model?
I initially thought of simulating some data from a triangular distribution with different modes that lie very close to min, max, or in between the min and max values. Then, suggest that depending on the risk tolerance, any of these three models can be selected. 
I am not sure if this method is justified or is there a better way to simulate some data that represent the unknown distribution. 
 A: If you know only the minimum and maximum, then I do not think that it is justified that either mode is at the minimum, at the maximum, or at the midrange. What follows, I do not think that using triangular distribution is appropriate in here unless you would select it's modes randomly from the interval between minimum and maximum. If you cannot assume anything besides the minimum and maximum, why not just use uniform distribution?
A: You can generate from triangular distribution as long as you disclose it to your users. This would be an assumption you made. If you were in banking or other regulated industry then you'd be asked to support your assumptions. Otherwise, it's your assumption and you're free to make it as long as your users are Ok with it.
However, if your question is whether it's a reasonable assumption given solely two data points, min and max, and absolutely no other prior knowledge, then the answer is NO. You need some other information to make this assumption, such as industry practices or prior research.
