When survey data is missing from all responses to a single question, the best method to address this issue would be to replace the missing data. This can be done using imputation techniques where appropriate, such as inserting a placeholder that indicates the data was not provided (e.g., 'N/A' or 'Not Answered'), or using statistical methods to estimate what the missing data likely would have been. This approach helps maintain the integrity of the dataset and allows for meaningful analysis, unlike removing duplicate or invalid data which are unrelated to the issue of missing data specific to a single question.