1
$\begingroup$

It is often pointed out that sample is an overloaded term in statistics and the sciences being supported by statistics. In my field (geological sciences) as in most other sciences, the process of collecting meaningful data is critical and discussions about the traps and pitfalls in that process talk about sampling. Not far down the road from that, particularly when lab results are back, conversations involving statisticians, data scientists, geomathematicians, GIS analysts and even normal geologists are likely to attempt to include multiple meanings of sample in the same sentence!

Q: Have any data scientists (or statisticians) found practical ways to communicate these different meanings?

One way is to always add soil, rock, statistical and so on before sample. But I was curious if there are any other approaches to effective communication that are in use.

Improve this question
$\endgroup$
5
  • 1
    $\begingroup$ I think that adding clarifying words to terms, ambiguous in a particular context (such as sample), is inevitable, if you want to communicate with the maximum clarity and prevent any miscommunication. $\endgroup$
    – Aleksandr Blekh
    Commented Jan 5, 2015 at 19:49
  • $\begingroup$ @AleksandrBlekh you mean writing something like "I analyzed a statistical sample of rock samples"? Honestly I'm not sure there is a distinction; just hierarchical data. $\endgroup$
    – shadowtalker
    Commented Jan 6, 2015 at 16:47
  • $\begingroup$ @ssdecontrol: Obviously, I didn't mean that. No need to combine together the same terms in the same sentence. What I meant is what you expressed by saying "rock samples". Using "rock" here makes it clear that geological, not statistical, meaning is implied. So, you can rewrite your phrase as follows: "I collected rock samples from <...> site (N=100). I used stratified random sampling for collection". No ambiguity, everybody's happy. $\endgroup$
    – Aleksandr Blekh
    Commented Jan 6, 2015 at 23:29
  • $\begingroup$ @AleksandrBlekh that's definitely more elegant. But I'm legitimately wondering if "statistical sample" is too cumbersome. The qualification should be bilateral IMO. Maybe "random sample" would be unambiguous enough $\endgroup$
    – shadowtalker
    Commented Jan 7, 2015 at 0:51
  • $\begingroup$ @ssdecontrol: I see your point and agree that "statistical sample" is an overkill. However, other than that particular word combination, I still stand by my opinion. I don't see how "sample" can qualify some other term, while the reverse is IMHO true in most cases. $\endgroup$
    – Aleksandr Blekh
    Commented Jan 7, 2015 at 1:00

1 Answer 1

Reset to default
1
$\begingroup$

"Sample" as a noun usually refers to a single data point. "Sample" as a verb is the act of extracting a subset of data points from some larger body (reality or a larger dataset). The only way to be less ambiguous is use more specific words than "data" or "sample".

Example:

Say you collect 1MM data points from four different sensors in the field giving you four sets of 250k data points. Say this data is too big for some demo of a model you're testing or an analysis you're running, so you pick 100k data points evenly split across the four sensors (giving four sets of 25k data points).

In this example, you're sampling twice. First, to gather your 1MM data points sampling from reality. Second, you sample again to decrease the size of your data set to something more manageable. 'Data' or 'sample' could refer to reality, the 1MM dataset, the 100k dataset, or any of the sensor-specific subsets. To make it less ambiguous, establish a unique name as soon as possible for each possible definition you'll be working with. ('reality' for the set of all possibly observed samples. 'the complete dataset', something derived from the source of the dataset, or even X for the full 1MM dataset. 'our trial dataset', or even Y for the small 100k dataset.

What you actually do comes down to context and what's appropriate for your intended audience, but the general answer is to use more specific words.

Improve this answer
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.