There are a number of uses and techniques for cutoff sampling (https://www.researchgate.net/publication/383976649_Cut-off_Cutoff_Quasi-Cutoff_and_Other_Nonprobability_Sampling_with_Inference), but the most accurate may be ratio modeling for Official Statistics when there is repeated/periodic sampling with an occasional census such as monthly economic/market reporting from a sample with an annual census for the same data items. This is described in https://www.researchgate.net/publication/362370770_Application_of_Efficient_Sampling_with_Prediction_for_Skewed_Data.
This may be most useful when the predictor data in a ratio model is from a census, and it is the same data item as the variable of interest in a sample. Each data item generally has its own ratio model, though with stratification, more than one model will be needed, or when borrowing strength, one model would be used across multiple data items. For such data items where one data item in a census is the predictor for that same data item in the sample, accuracy is generally high, as may be found in the case of electric power data and natural gas receipts at the US Energy Information Administration. Testing is also feasible. (If more than one data item in the census might contribute to predicting one data item in the sample, this has also been found to do well when necessary.)
It could be helpful to at least include the elementary case of each data item in a sample predicted by the same data item in a previous census. Sample size information for this is shown in https://www.researchgate.net/publication/261947825_Projected_Variance_for_the_Model-based_Classical_Ratio_Estimator_Estimating_Sample_Size_Requirements.
Because a sample may have very large values in the corresponding census for some data items but not as large for others, it may be a cutoff sample for some data items and a near-cutoff or quasi-cutoff sample for other data items.
Variance estimation in this case is simple. See https://www.researchgate.net/publication/359204410_Variance_of_the_Prediction_Error_for_Totals_Under_Cutoff_or_Quasi-Cutoff_Sampling.
In summary:
This might be most helpful for periodic economic sample surveys with an occasional census of the data items of interest.
See https://www.researchgate.net/publication/303496276_When_and_How_to_Use_Cutoff_Sampling_with_Prediction.
(For more, see the following on small area prediction: https://www.researchgate.net/publication/262066356_Quasi-Cutoff_Sampling_and_Simple_Small_Area_Estimation_with_Nonresponse.)
Also, see https://www.researchgate.net/profile/James-Knaub/research.
There are a number of uses and techniques for cutoff sampling (https://www.researchgate.net/publication/383976649_Cut-off_Cutoff_Quasi-Cutoff_and_Other_Nonprobability_Sampling_with_Inference), but the most accurate may be ratio modeling for Official Statistics when there is repeated/periodic sampling with an occasional census such as monthly economic/market reporting from a sample with an annual census for the same data items. This is described in https://www.researchgate.net/publication/362370770_Application_of_Efficient_Sampling_with_Prediction_for_Skewed_Data.
This may be most useful when the predictor data in a ratio model is from a census, and it is the same data item as the variable of interest in a sample. Each data item generally has its own ratio model, though with stratification, more than one model will be needed, or when borrowing strength, one model would be used across multiple data items. For such data items where one data item in a census is the predictor for that same data item in the sample, accuracy is generally high, as may be found in the case of electric power data and natural gas receipts at the US Energy Information Administration. Testing is also feasible. (If more than one data item in the census might contribute to predicting one data item in the sample, this has also been found to do well when necessary.)
It could be helpful to at least include the elementary case of each data item in a sample predicted by the same data item in a previous census. Sample size information for this is shown in https://www.researchgate.net/publication/261947825_Projected_Variance_for_the_Model-based_Classical_Ratio_Estimator_Estimating_Sample_Size_Requirements.
Because a sample may have very large values in the corresponding census for some data items but not as large for others, it may be a cutoff sample for some data items and a near-cutoff or quasi-cutoff sample for other data items.
Variance estimation in this case is simple. See https://www.researchgate.net/publication/359204410_Variance_of_the_Prediction_Error_for_Totals_Under_Cutoff_or_Quasi-Cutoff_Sampling.
In summary:
This might be most helpful for periodic economic sample surveys with an occasional census of the data items of interest.
See https://www.researchgate.net/publication/303496276_When_and_How_to_Use_Cutoff_Sampling_with_Prediction.
(For more, see the following on small area prediction: https://www.researchgate.net/publication/262066356_Quasi-Cutoff_Sampling_and_Simple_Small_Area_Estimation_with_Nonresponse.)
Also, see https://www.researchgate.net/profile/James-Knaub/research.