Limits...
Constrained Dual Scaling for Detecting Response Styles in Categorical Data.

Schoonees PC, van de Velden M, Groenen PJ - Psychometrika (2015)

Bottom Line: Response styles occur when respondents use rating scales differently for reasons not related to the questions, often biasing results.A spline-based constrained version of DS is devised which can detect the presence of four prominent types of response styles, and is extended to allow for multiple response styles.An alternating nonnegative least squares algorithm is devised for estimating the parameters.

View Article: PubMed Central - PubMed

Affiliation: Econometric Institute, Erasmus University Rotterdam, Rotterdam, The Netherlands. schoonees@gmail.com.

ABSTRACT
Dual scaling (DS) is a multivariate exploratory method equivalent to correspondence analysis when analysing contingency tables. However, for the analysis of rating data, different proposals appear in the DS and correspondence analysis literature. It is shown here that a peculiarity of the DS method can be exploited to detect differences in response styles. Response styles occur when respondents use rating scales differently for reasons not related to the questions, often biasing results. A spline-based constrained version of DS is devised which can detect the presence of four prominent types of response styles, and is extended to allow for multiple response styles. An alternating nonnegative least squares algorithm is devised for estimating the parameters. The new method is appraised both by simulation studies and an empirical application.

Show MeSH
The effect of response styles on the underlying uncorrelated objects: estimated Pearson correlations before and after contamination, as well as after cleaning the data. The number of rating categories is  for (a)–(c) and  for (d)–(f), with  items in all cases.
© Copyright Policy - OpenAccess
Related In: Results  -  Collection


getmorefigures.php?uid=PMC4644217&req=5

Fig5: The effect of response styles on the underlying uncorrelated objects: estimated Pearson correlations before and after contamination, as well as after cleaning the data. The number of rating categories is for (a)–(c) and for (d)–(f), with items in all cases.

Mentions: The simulation model of Section 4.1 assumes that, given the expected value of the object scores , the objects are independently distributed as truncated normal distributions. Although the true correlation matrix between the objects thus is the identity matrix , the observed correlations after the response style contamination is often inflated. To show improvement, the cleaned data derived as in Section 3.6 should have correlations resembling independence more closely. A visual example is given in Figure 5 for simulated data ( similar to the conditions used in Tables 2 and 3), where the colours indicate the magnitude of the Pearson correlations. It is evident that the response styles artificially inflate the correlations. When , the cleaned data to some extent succeeds in removing the spurious correlations, but when the situation is not much improved.Fig. 5


Constrained Dual Scaling for Detecting Response Styles in Categorical Data.

Schoonees PC, van de Velden M, Groenen PJ - Psychometrika (2015)

The effect of response styles on the underlying uncorrelated objects: estimated Pearson correlations before and after contamination, as well as after cleaning the data. The number of rating categories is  for (a)–(c) and  for (d)–(f), with  items in all cases.
© Copyright Policy - OpenAccess
Related In: Results  -  Collection

Show All Figures
getmorefigures.php?uid=PMC4644217&req=5

Fig5: The effect of response styles on the underlying uncorrelated objects: estimated Pearson correlations before and after contamination, as well as after cleaning the data. The number of rating categories is for (a)–(c) and for (d)–(f), with items in all cases.
Mentions: The simulation model of Section 4.1 assumes that, given the expected value of the object scores , the objects are independently distributed as truncated normal distributions. Although the true correlation matrix between the objects thus is the identity matrix , the observed correlations after the response style contamination is often inflated. To show improvement, the cleaned data derived as in Section 3.6 should have correlations resembling independence more closely. A visual example is given in Figure 5 for simulated data ( similar to the conditions used in Tables 2 and 3), where the colours indicate the magnitude of the Pearson correlations. It is evident that the response styles artificially inflate the correlations. When , the cleaned data to some extent succeeds in removing the spurious correlations, but when the situation is not much improved.Fig. 5

Bottom Line: Response styles occur when respondents use rating scales differently for reasons not related to the questions, often biasing results.A spline-based constrained version of DS is devised which can detect the presence of four prominent types of response styles, and is extended to allow for multiple response styles.An alternating nonnegative least squares algorithm is devised for estimating the parameters.

View Article: PubMed Central - PubMed

Affiliation: Econometric Institute, Erasmus University Rotterdam, Rotterdam, The Netherlands. schoonees@gmail.com.

ABSTRACT
Dual scaling (DS) is a multivariate exploratory method equivalent to correspondence analysis when analysing contingency tables. However, for the analysis of rating data, different proposals appear in the DS and correspondence analysis literature. It is shown here that a peculiarity of the DS method can be exploited to detect differences in response styles. Response styles occur when respondents use rating scales differently for reasons not related to the questions, often biasing results. A spline-based constrained version of DS is devised which can detect the presence of four prominent types of response styles, and is extended to allow for multiple response styles. An alternating nonnegative least squares algorithm is devised for estimating the parameters. The new method is appraised both by simulation studies and an empirical application.

Show MeSH