CAPÍTULO I. EL APORTE DE LOS NUEVOS ENFOQUES NO POSITIVISTAS AL ANÁLISIS DEL CAMBIO EN POLÍTICAS PÚBLICAS
1.1. EL ANÁLISIS DEL CAMBIO EN POLÍTICAS PÚBLICAS
1.1.2. Las críticas a los enfoques clásicos. El desarrollo de nuevos enfoques sobre el cambio de políticas
External validity or generalisability of research outcomes
As described in the previous section, for this thesis research, cases were non-randomly sampled, hence seeking statistical significance of results in the wider population, obtained from a small sample and non-randomly sampled case-studies, is not possible (Silverman, 2010, Robson, 2011, Saunders, 2009, Flyvbjerg, 2006) neither a purpose of this thesis research. However, purposively selected case-studies were selected based on maximising understanding and learning and can give meaningful and detailed insights and refinements for the testing and building of methodology and theory, which can be transferable and useful to the wider population where applied in similar conditions and with similar purpose (Silverman, 2010, Flyvbjerg, 2006, Stake, 1995). Generalisability beyond the case studied does not need to be based on statistical inferences (Robson, 2011) and "may be thought of as the development of theory which helps in understanding other cases or situations" (Robson, 2011) and any trends, patterns and findings "are the basis for generalisation" (Robson, 2011). Case-studies can also highlight observed phenomena and lead to refinement or modification of pre-existing generalisations (Stake, 1995). This was done by for example Lowe (2007) who demonstrated the presence of a thermal bypass in millions of cavity-walls based on a single case-study. Indeed, findings based on purposive case-study data collection may lead to generalisations to inform more about typical case-studies (Saunders, 2009) and may support the refinement and modification of models and theories (Stake, 1995), which might also apply in other cases with similar characteristics. Case-studies could also highlight interesting findings about research methods and efficacy of proposed interventions (Silverman, 2010), which may be applicable elsewhere and support future research. Finally, case-studies can also be useful to highlight certain trends in the data collected (Robson, 2011) and to generate and test hypotheses (Flyvbjerg, 2006), as undertaken in this thesis. Hence while findings are specific to the particular case-studies and not representative of the wider population, they may highlight methodological and theory implications as well as highlight generalisable trends and patterns and will inform future research and support future studies.
The high-resolution heat-flow measurements undertaken in the Salford Energy House environmental chamber (STUDY 2, Chapter 4), are not representative of the wider pre-1919 housing population. Additionally, and as discussed in Section 4.3., the Salford EH has been reconstructed in an environmental chamber and there are several variables which are not replicated from an actual house.
With regards to the field study (STUDY 4, Chapters 5 and 6), the experimental research design was designed to be as representative as possible of an occupied dwelling, for example by replicating occupied dwelling heating patterns. However, each research design has
limitations, such as not replicating the effect of uninsulated radiator pipes in the void, taking measurements in an unfurnished room, etc. Additionally, the field study house is almost certainly not representative of the entire pre-1919 housing stock in the UK. In particular, the field study had over-site concrete, 3 sleeper walls with limited openings in between floor void sections; the airbricks were located within the joist spacings and the total ground floor area was small. This study conducted cross-sectional, not longitudinal work due to access limitations, hence this research excludes trends over time. More dwellings need to be studied to assess the effect of differences in the housing stock and to understand if any observed trends exist for floors with different characteristics and over a longer time-period. Despite the above caveats, several of the research findings were transferable and applicable to the wider population. A summary of the issues are described in Table 14.
below.
Internal validity refers to replicability and to the accuracy of the findings and whether they capture their intended target and in the case of interventions, whether findings are due to the chosen intervention (i.e. causation) and not due to other influences (Robson, 2011). To support internal validity, hypotheses were tested and confounding variables were controlled where possible (such as heating schedule, using the same sensors in the same locations between interventions, minimising researcher and occupant influences etc.). Additionally, a detailed field diary was made and research design is reported transparently and in-depth.
General concerns How to address these concerns?
Replicability of research (Internal validity)
Hypotheses testing; control confounding variables as much as possible (such as heating schedule, sensor replacement between interventions, etc.), other independent variables were measured to qualitatively assess the impact of changing environmental conditions on observed heat-flow.
Additionally, the research method and data collection procedures were noted in a detailed field diary during the monitoring process; regular (digital) archiving of field notes was undertaken. Calibration checks were undertaken where possible. Research design is transparent and reported in-depth.
Generalisability outcomes (External validity)
The purposive case-study approach cannot provide population-wide inference (Saunders, 2009) but was used to generate and test hypotheses (Flyvbjerg, 2006), from which trends, patterns and findings can be "the basis for generalisation" (Robson, 2011) and which can be transferable and useful to the wider population where applied in similar conditions and with similar purpose (Silverman, 2010, Flyvbjerg, 2006, Stake, 1995).
Table 14. Summary table of general concerns and how to address these
3.5. Summary
This chapter introduced in-situ U-value measuring and analysis techniques and associated uncertainties in detail, alongside raising conceptual issues of measurement location and difficulties of low resolution heat-flux point measurements as a comparison to whole floor U-value models. These issues are further explored in the following chapter, based on in-situ measurements in the Salford Energy House. This chapter also set out the main studies and sampling strategies conducted in this thesis research and associated ethical and general concerns, including generalisability of findings based on case-study research.