XLSTAT: Raking a Survey Sample

 

Dataset for Example

The (attached, simulated) data correspond to a survey based on 200 employees of a company on their satisfaction at work.

Two auxiliary variables were included in the survey: gender (1: male, 2: female) and age representing the accumulated experience in the company in three categories (1: 1 to 14 years, 2: 15 to 29, 3: 30 to 45 years). The proportions of these variables in the entire company are known (marginal control totals).

There are 10000 employees in this company.

We are interested in finding raking weights that can be applied to our survey sample in order to obtain similar proportions for the modalities of the auxiliary variables, in the survey sample and in the population (Deming and Stephan, 1940).

 

Setting Up

• Open XLSTAT.

• Click on the Marketing/Marketing tools menu and select the Raking survey feature.

In the General tab:

• Select the auxiliary variables (gender and age).

• The marginal control totals have to be selected all together in the same order as the data to be raked on (one column for each variable; one row for each modality).

• Each column has to sum to the same value (here 10000).

• As we selected the column title for the variables, left the option Variable labels activated.

In the Options tab:

• Select  Raking Ratio as estimation method.

• Click on the OK button to launch the computations.

• The results are displayed on a new Excel sheet.

 

Interpreting the Results

The first results displayed by XLSTAT are the basic statistics associated with the auxiliary variables before raking.

The second table contains the final weights for each observation, the initial auxiliary variables and the weights ratios (final weights / initial weights) (see below for observations 1 to 13).

The basic statistics after raking are also displayed. We can see that using the obtained weights, the statistics are equal in the survey sample and in the population.