5.3 Estimation of ratios of population means

Suppose we know the population mean \(\mu_x\) of an auxiliary variable \(x\), and it is associated with the variable of interest, \(y\). To estimate the population means of the target variable \(\mu_y\), intuitively, we could first estimate the ratio of their population means, $R_{y,x} $ and multiply it to \(\mu_x\), such that \(\hat\mu_{yR}= \hat R_{y,x} \mu_x\), where the subscript “\(R\)” denotes “Ratios”, see (Wu and Thompson 2020). In this chapter, we demonstrate how to estimate the ratio between the population means of two variables. We created a variable \(\texttt{HWT_DHT_M_TRM_sq}\), which is the square of the self-reported height (in meters). The estimate of the ratio between the population mean of \(\texttt{HWT_WGHT_KG_TRM}\) and that of \(\texttt{HWT_DHT_M_TRM_sq}\) is computed as

\[ \frac{\sum_{i\in \mathcal{S}} w_i (HWT\_WGHT\_KG\_TRM_i) }{\sum_{i\in \mathcal{S}} w_i (HWT\_DHT\_M\_TRM\_sq_i )}, \] where \(w_i\) is the survey weight and \(\mathcal{S}\) is the set of sampled units.

R

svyratio(numerator = ~HWT_WGHT_KG_TRM, denominator = ~HWT_DHT_M_TRM_sq, 
         design = CLSA.design )

SAS

PROC SURVEYMEANS data = CLSAData   ratio;   
VAR HWT_WGHT_KG_TRM HWT_DHT_M_TRM_sq;                    
RATIO  HWT_WGHT_KG_TRM / HWT_DHT_M_TRM_sq ;
STRATA GEOSTRAT_TRM;                       
WEIGHT WGHTS_INFLATION_TRM;                                   
RUN;   

SPSS

Click “Analyze” \(\rightarrow\) “Complex Samples” \(\rightarrow\) “Ratios…” \(\rightarrow\) Browse and select the survey design file, “\(\texttt{CLSADesign.csaplan}\)\(\rightarrow\) Click “\(\texttt{Continue}\)” and enter variables “\(\texttt{HWT_WGHT_KG_TRM}\)” and “\(\texttt{HWT_DHT_M_TRM_sq}\)” to the the “Numerators” and “Denominator” panels, respectively \(\rightarrow\) Click “Statistics…” \(\rightarrow\) Select “Standard error” \(\rightarrow\) Click “Continue” \(\rightarrow\) Click “OK”.

Stata

svy linearized : ratio (HWT_WGHT_KG_TRM / HWT_DHT_M_TRM_sq)
Result comparison
Estimate R SAS SPSS Stata
Estimate 27.6076 27.6076 27.6076 27.6076
SE 0.3177 0.3177 0.3177 0.3177

Reference

Wu, Changbao, and Mary E. Thompson. 2020. Sampling Theory and Practice. Cham, Switzerland: Springer. https://doi.org/10.1007/978-3-030-44246-0.