Please, I will perform a systematic observation (naturalistic observation) over a group of users. As my observation will be completely non-obtrusive, I can not gather detailed data directly on each individual observed. My question is, how valid can be a survey conducted on the sample as whole (the large group of individuals present at a given area), but not being exactly the individuals observed? Thanks!

  • Might this be a question for Cross Validated?: stats.stackexchange.com
    – wgmcgowan
    Commented May 19, 2019 at 19:54
  • I did not know about Cross Validated. Thank you.
    – Vinicivs
    Commented May 20, 2019 at 3:28

1 Answer 1


I assume that the set of individuals surveyed is not the entire population of interest, but only a sample. If it is the population, then, of course it is valid. If it is a sample it can be valid or not depending on the usual factors such as sample size and assumptions of randomness, etc.

The observations can be valid or not (as giving information about the population) also, depending on how conducted.

But what you wind up with is two pieces of evidence about the population and they might agree or not. But they need separate analysis as well as argument about what you have overall based on the two separate experiments.

But, I also caution you about doing the observations yourself. It is a dangerous practice and can easily lead to observational error. You might "see" what you "want to see", introducing observer bias. It is much better to use an uninvolved observer that you train to record what is necessary, rather than doing it yourself. Such bias can be entirely unintentional and still be present.

  • Thank's @Buffy. From your response, my question could be formulated like this: considering same population but different samples, can I do correlation analysis with data collected from survey and data collected from observation? How valid can be this correlation among different samples of the same population? Also, regarding observation, although I know that participant/non-participant observation could give more valuable and meaningful data, I am trying to reduce bias through a strictly and systematic observation method (behaviour mapping). Feel free to give me more insights about that.
    – Vinicivs
    Commented May 20, 2019 at 3:25

Not the answer you're looking for? Browse other questions tagged .