Using PLS to analyze sales data

[hspotts]

Hi!

Have assembled a variety of data related to satisfaction and media/social media activity for a set of 48 firms. Have 18 quarters of data for each firm and am interested in looking at effects on sales performance. Is it possible to use SmartPLS to analyze this data? All variables are single item measurements.

Thank you.

[jmbecker]

If all variables have single item measurement it would be possible, but not convenient to use PLS. I would rather use R or STATA.

[hspotts]

We are interested in relationships among several different marketing inputs, consumer behavior and sales measures. Some of the data are skewed because they were either measured using top box scores or had some valid zeros (no ad spending). If we were not interested in time period effects and mean centered the data over time periods for each firm, would it be possible to use SmartPLS for the analysis? I know there are time series approaches that can be used, but we are not interested in the time series component.

[jmbecker]

PLS is a regression based technique and thus you can use it for quite many purposes. It can also be used for your type of analysis, but you need some experience in modeling and interpreting the coefficients. If you have only single item measures and no latent variables it is currently not very convenient to use SmartPLS, because of the inherent standardization of data in the PLS algorithm. We will provide additional simple path modeling algorithms in the close future that circumvent this "problem". With those algorithms it might be more interesting to model you data. Until then, I would rather use a more econometric focused software if you are not having any multi-item measures (latent variables).