Hi,
I have come across research in which all population were approached to participate in the survey via email. What would we call this approach?
Thank you.
That is attempting to do a census.
However, there are likely to be nonresponses which are not at random. That is, the mean response for the responders is likely to be different from what the nonresponders would have given for a particular quantitative item on the survey, were they have been inclined and able to give reliable responses. Thus, imputation or else weighting by categories would be needed which take this into account. If complete regressor data were available, then "prediction" (estimation for a random variable, not forecasting) could be done, and variances estimated. Bias may be avoided to a large degree, with a good model, the bias coming from the fact that no model is exactly 'correct.'
In fact, the entire distribution of responses for a given item could be very different from the true population. That may differ by item, and similarly, for qualitative information, there could be differences as well.
For a quantitative, say continuous data item, if there were a previous census that has been completed, then that could provide the best predictor data for a ratio model, as in the environmental data example in https://www.researchgate.net/publication/362370770_Application_of_Efficient_Sampling_with_Prediction_for_Skewed_Data.
You could think of the responses for each data item as a nonprobability sample, and you need to impute, or weight, to fill in for nonresponses. Without good auxiliary/predictor/covariance data, that could be very inaccurate.
PS - The statistician, George Box, famously claimed that models are incorrect, but can sometimes be "useful," to remind us, I think, that you can never model exactly. However, when you can use the same data item in a previous census as a size measure, say in a ratio model, this can often work extremely well. (See Cochran, W.G.(1953), Sampling Techniques, 1st ed, John Wiley & Sons, pages 205-206.)
not dissimilar to a electoral voting system you could look up a political site and see what they call a voting poll where all members are approached
- Badges
- Science topic
- Similar topics
- Methodology
- Research Methodology
- Research Papers