Before you start to collect data, it is essential to understand the sample frame, sample techniques, and possible sampling errors. Even with careful survey planning, the sample you cap in does not tally with the expected sample size you aimed to meet.
It can be because of a variety of factors like time limitations, budget constraints, improper sample frame, or high non-response. To mitigate these sample disparities, a survey weighting method is employed.
What is the weighting concept?
OvationMR has experts with experience in data weighting principals. Weighting is a statistical technique where the dataset is manipulated via calculation to bring the targeted population in line.
The main difference between initial sample structure and weighting is the latter is applied after data collection. It allows the researchers to fix issues that happened during the data collection process. So weighting is also called ‘Post Stratification’. Pre stratification is applied to even out the sample before data collection.
Weighting data
Pros
- Allows making the data set right, so the results that represent the targeted population are near to accurate.
- Makes sure that opinions of unreachable demographic groups are considered in the final result.
- The effects of data collection challenges or inherent biases [associated with the survey method used] are reduced.
Cons
- Possibly overrepresent several or a single person’s view, who may never reflect the entire demographic group correctly.
- Probably make findings more erratic as standard deviation increases.
- Can involuntarily introduce extra biases within the dataset.
To reduce the data weighting impact, experts recommend using few variables for weighting. The increase in weighting variables the greater the complexity, so minimizes weight sizes. The general rule is not to weigh a survey participant less than .5 and not more than 2.0.
Common weighting methods used in the survey
Cell-based weighting
When you are aware of the sample size then the cell-based weighting method can be used. For example, if your sample includes 100 females aged 20 to 30 and 80 males aged 25 to 35, but you wanted to have 90 females and 110 males then you can use cell-based weights.
RIM [Raking] weighting
Random Iterative Method or RIM is a little complex way that is used when multiple variables cause confusion on how to interlock them. For example, you need 120 females and 150 people aged 24 to 36 but are not aware of the number of females aged 24 to 36 is needed. In raking, the sample is balanced on a single variable i.e. gender and then another variable i.e. age is applied.
If adjustment of one variable impacts another variable then extra corrections are made until researchers achieve a balanced sample. For public opinion surveys, raking is an accepted weighting method because it allows the use of multiple variables and flexibility to adjust each variable. Using a specific software raking method is quickly performed.
Other methods that can be applied are matching, propensity weighting, and logistic regression modeling. For more in-depth clarification a combination of these different weighting methods is used. Contact professionals to learn more about the weight application and different weighting methods.
