Hello everybody,
a colleague and me, we are working on an study that aims at analyzing (and rating) climate change strategies of companies and the interlinkages with companies' actual carbon performance.
We have about 40-50 companies from two sectors: energy-intensive (steel & cement; about 15 cases) and automotive industry (about 30 cases).
We developed a scheme for assessing the climate change strategy of these companies. It consists of 10 different measures that companies can implement in response to climate change (e.g. product/process improvement, emission trading, etc.). Based on publicly available data (CDP database, sustainability reports, etc.) we rated each company in each of these 10 categories/measure with a score from 0-4 (doing nothing-being highly active).
Additionally, we have calculated carbon performance indicators including carbon intensity (CO2 per USD sales) for each company.
We have done these two analyses (strategy and performance assessment) for 2008 and 2013.
Now my question: which (statistical) method would/could you recommend for
...comparing the companies/sectors
...comparing the companies' performance between the two points in time
...analyzing the relationship between strategy and carbon performance?
Thanks in advance for your help! It's much appreciated.
Best regards,
Matthias
Use longitudinal study
Use PLS SEM because, PLS SEM not assume normality of data and work pretty well for small samples, it do bootstrapping (I can provide provide you a paper in this sense) if you use CB SEM you must have a 200 cases simple.
Use regression, but make sure about the proper normality, linearity, homocedasticity, skewness and kurtosis of data.
I would suggest a repeated measures MANOVA based on your description, with the time variable being the two years in your study which is longitudinal. To run the analysis in R, I would suggest using the car package. I have attached instructions for the analysis and package installation. This should handle the across time question. You can do this by sectors or by company. The relationship between strategy and carbon performance may require a SEM. Note that all of these tests have a normality assumption so Carlos Monge's comment about PLS SEM may be appropriate. I have attached a paper on SEM modeling to get you started. The RobPLS talk gives a start into nonnormalty. Transformations may be helpful if your data isn't normal. RMultiVar gives you a start on plotting multivariate data which you will want to do. Best wishes.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Applied Statistics