Page 219 - Materials Chemistry, Second Edition

10.5 Challenges in the application of MADM for LCSA 217
top performers (Kalbar et al., 2017a). Finally, indicator uncertainty involves selection of indi-
cators in a study that are irrelevant or incomplete (Heijungs and Huijbregts, 2004). Addition-
ally, uncertainties could also be associated with the framing of the problem, selection of
method for aggregation, and levels of selected attributes (Scholten et al., 2015). There is no
clearly documented way to fully remove the uncertainty as the indicator selection is a sub-
jective process and depends on the domain(s) knowledge of involved researcher(s).
Tzeng and Huang (2011) suggest that before proceeding with MADM analysis, data must
be put into a histogram and their distribution with standard deviation should be checked.
If the distribution is nonnormal and standard deviation is significant, then sensitivity analysis
is mandatory during MADM analysis. Sensitivity analysis must precede uncertainty analysis
(Kleijnen, 1994).

10.5.7 Interpretation of the results
Results from an analysis in MCDA typically involves ranking of alternatives concerning a
specific set of attributes. The results are obtained in the form of an aggregated index and need
further interpretation for deriving correct decision support. For example, Figueira et al. (2005)
and Munda (2005) conducted studies on four different cities using a distance-based method
such as TOPSIS. The studies suggested that results from MADM analysis cannot blindly be
relied upon. Even if different types of aggregation schemes are used and still the results are
not robust, then reconsideration of areas related to uncertainties mentioned in Section 10.5.6
must be completed. Therefore, it is recommended that in MADM analysis, robustness of the
decision process is more critical compared to the final solution.
Similarly, LCSA also has a major challenge in its interpretation of results, where the
integration of three different tools (LCA, LCC, and SLCA) is required to produce a collective
result (Hannouf and Assefa, 2017). Zhang and Haapala (2015) suggested the use of MADM
approaches as an efficient way of developing frameworks to integrate the tools and interpret
combined results. Zampori et al. (2016) provided general guidelines to interpret results,
in which identification of significant issues can be achieved by the use of MADM methods.
Additionally, the study also recommended conducting thorough checks like completeness of
inventory data, sensitivity analysis to assess reliability of results, and consistency check of
methods and assumptions. There has not been sufficient work on the interpretation of results
in MADM integrated with LCSA. One of the efforts for interpretation of MADM results is
using radar diagrams, as demonstrated by Kalbar et al. (2012).
The results of ranking in LCSA based on MADM are an aggregated score, i.e., a single value
for each indicator. Considering all the methodological choices, data uncertainties, effects
of weights, and MADM methods limitations (e.g., rank reversal), unless the topmost ranked
alternatives have a significant difference in the score from the second most alternative, that
alternative cannot be concluded as the best performing one. For example, Kalbar et al. (2016)
implemented an approach wherein such cases, the top two to three alternatives having almost
equal scores will be concluded as most-preferred alternatives.
In a reallife situation, as best practice, it is recommended to apply multiple MADM
methods for the given problem with different weighting schemes reflecting the priorities
of stakeholders. The alternatives that are frequently ranked as topmost can be concluded
as the most preferred ones.