It seems that you are hitting against some limitations of keyword searches, but that could be seen as a good thing because keyword searches should not be the primary strategy for literature searches. Keyword searches are often (rightly) the first step of the search process. However, we must realize that no matter how carefully we craft our keywords, such searches will necessary miss many important articles.
I propose the following search procedure that should help you find what you're looking for.
First, you need to adopt a recursive approach for the keyword search. So, you could begin by searching directly for your desired keywords without worrying that you will miss some articles. So, based on your sample string above, you can begin with something like this:
"missing data" OR "sampled data" OR "imperfect data" OR "unknown data" OR "partially observed data"
Suppose, as an example, this comes up with 100 results. Read the abstracts and filter to the articles that are truly relevant for your purpose. Suppose you have 50. Examine the full text of these articles; suppose you find that only 20 are truly pertinent. Then you can start reading these 20.
As you read these pertinent articles, pay attention to other synonyms of "missing data" that they might have. Suppose, for example, you find that some of them refer to terms like "missing at random" or "missing not at random". After reading these articles and collecting the synonyms, you should go back and do another round of searching with these new keywords. Again, filter abstracts, find the new pertinent articles, read them and gather any new synonyms. After a few rounds of this, let's say that you have identified 30 articles that your keyword searching can find for you. Now you're ready for round two.
The second major type of searching is backward- and forward-citation searching. In both cases, you start with the articles that you have found that are really exactly what you are looking for. Let's suppose that out of the 30 articles from your keyword searching, only 10 are perfectly what you want. You will carry out the following steps for each of these 10. (In principle, you could do it for all 30, but then the procedure gets really tedious and might not add that much.)
First the backward citation search. For a pertinent article, read the titles of all the articles in the cited references list. Take note of any title that seems interesting. Then look up all such interesting titles, read their abstracts, and then collect those that are relevant for full-text reading. When you find any that is indeed pertinent, you do the same thing with it: read the references list, identify interesting articles and look them up. You should keep on doing this until you find no further interesting articles that you have not already included. This procedure is called a backward citation search because all such articles that you find are published backwards in time relative to each article that you are focusing on.
Next, for each truly pertinent article, you should also do a forward citation search. For this, search for the article in Google Scholar. Right underneath the article, you should see a tiny link that says something like "Cited by X" where "X" is the number of articles that have cited that article. Click on that link and then Google Scholar will list for you all the articles that have cited that article. Browse through those titles (at least the first 30 if there are too many). If any title looks interesting, then add it to your list of articles to examine abstracts, full text, etc. This procedure is called a forward citation search because all articles that you find are published forwards in time (that is, later than) relative to the focal article of interest.
Whenever you identify a new article that is truly right-smack on your topic of interest, you should carry out a full backwards and forwards citation search on it. If you follow this procedure, you will likely find several additional articles that your keyword search had not identified.
Hopefully, a more thorough search procedure like this might meet your needs.
