I am doing some statistical studies and I would appreciate some guidance to some approximation techniques when not all data is available. I have a model that takes certain input parameters (discrete, deterministic) and all of these are defined within a known interval.
What I would like to do is approximate the effect from every data point or variable, by validating using a complete case model. However, I am now sure how to go about doing this. I have read about some approximation techniques such as MCAR, MAR, NMAR etc. as well as imputation techniques and linear regression. But none of these seem to fit my problem, because these assume that data is missing from some kind of observation sample or distribution. That the data that is missing is somehow correlated to some dependent variable.
But in my case, the data that is missing is completely independent from any other data point, and the parameter values themselves contribute to a "total sum" (hence, the total sum can be defined to be within a certain interval for sure if it is known what is missing). But I would appreciate guidance to topics around these kinds of problems. Is regression analysis an appropriate methodology?
Thanks in advice!