As a result, measures of scientific production (which includes these socio-technical features) are a more relevant indication of likelihood to adopt than measures of scientific progress, although they could also indicate a potentially contentious or controversial topic that is generating lots of different opinions. However, controversy does not prevent adoption, and in some cases may accelerate adoption
Methodology
Statistical comparisons of time series
Problem:
Comparison of dissimilar bibliometric technology profiles to investigate whether literature based technology substitution groupings may be determined from separate measures of science and technology
Dissimilarity between technology time series:
- Time series based on different number of observations (e.g. different timespans)
- Time series may be out of phase
- Time series at different stages through the Technology Life Cycle
- Time series may be representative of dissimilar industries
- Time series may repeatedly move between different stages of the Technology Life Cycle
- Time series may be subject to long-term and shorter term cyclic trends
Potential time series classification methods:
- Supervised learning (i.e. training examples all have known classification labels) \cite{lin2012pattern}
- Semi-supervised learning (i.e. training examples have both known and unknown classification labels) \cite{lin2012pattern}
- Unsupervised learning (i.e. data has unknown classification labels, e.g. clustering) \cite{lin2012pattern}
Pre-processing of time series:
Statistical significance testing:
- Chi-square testing is best suited to confusion matrices with cell values all greater than 5: Fisher’s exact test is more appropriate for small sample sizes
- Histograms useful for determining most frequently occurring individual factor in ‘bootstrapping’ processes, but cannot indicate what combination of factors would work best together
- Fisher’s exact test can determine the significance of outcomes for random samples, but not necessarily the ranking of the most robust predictors (cross-validation approaches are required for this)
- Fisher’s exact test assumes that samples are taken at random from a population. In this study, technologies have been deliberately selected based on their observed characteristics – as such, Fisher’s exact test cannot be used to reject the null hypothesis, but can be used to determine the likely probability of predictor groupings
Time series pattern recognition techniques
Examples of common time series pattern recognition techniques (not exhaustive):