Page 69 - Contributed Paper Session (CPS) - Volume 8

CPS2187 Lukasz Widla-Domaradzki
The solution proposed here is based on anchoring dependent SEM models
into the basic one to obtain better analytical result. Through this method we
are able to build smaller models independently and merge them analytically
at the later stage.

2. Methodology
Anchored SEM is the analytical tool that allows to analytically create a
connection between at least two independent SEM models drawn from the
same sample. Those models should be – at the later stage – part of one
complex solution. It is possible to include more than two models in this
procedure, however, in this paper I focus on the simple solution when only
two models are merged. Another – mentioned above – constraint is models
should be dependent. In other words, one of the models should be able to
explain a part variability of the second one. If this condition is not satisfied (for
example with models which are related, but there is no dependency between
them) Anchored SEM can’t be used.
In the analysis presented below I worked with two independent SEM
models drawn from the same sample. Sample for this study was obtained from
the population of over 100 thousand of Polish enterprises. One of the aims of
the study was to analytically contribute to the large and complex indicator of
4
Polish Enterprises Innovation Index. Theory for the proposed index is well
described, so the SEM might be used.
The database used for presented analysis was coming from a pilot study
and contained 1327 cases. Difficult and multi-layered model that was needed
to construct a nationwide index demands bigger sample, but I attempted to
build a preliminary model on limited data and revalidate it after several waves
of the study become available. In order to do this, I had to develop some kind
of shrinking or folding models to connect some of the SEM latent variables to
each other. As a first step of this exercise, I prepared two connected (at the
theory level) models A2 and A3 (see Fig 1). Model A2 contained variables
connected with innovative infrastructure (such as: “level of process
automation”; “software for supporting ERP (enterprise resource planning)” or
“the number of employees who are responsible for developing innovations”
etc.), while model A3 included variables connected with innovative
management (such as level of agreement with the statements: “innovations
are supported by company managers”, “there is an innovation management
system”, “there are processes that allow us to effectively manage the
development of a new product” etc.). In theory innovative infrastructure is a

4 Theory of innovation is mainly based on expectancy-value theory first proposed by John
William Atkinson and expanded for the field of innovation studies by Martin Fishbein (in
Stephen W. Littlejohn’s “Theories of Human Communication”)
58 | I S I W S C 2 0 1 9