2.1.2. Statistical Links between the Results

Once we obtained this value, its links with the morphological aspects and the climatic circumstances of the dwellings, as well as the technical characteristics of their building materials, were analyzed, in order to better understand its functioning. By means of statistical graphs, it was possible to link morphological characteristics, that is to say qualitative data to quantitative data, such as the thermal transmittance or the wind speed. Therefore, in order to ascertain which factors influence the value of Δh, the statistics software PAST v3.25 [71–73] was used.

By means of it, a canonical correspondence analysis was carried out, that is to say a correspondence analysis based on a site/species matrix, in which values for one or more environmental variables are assigned to each site/specie.

The ordination axes of the resulting graph show the values of the combinations of those variables. This type of analysis is a direct gradient analysis, in which the gradient in environmental variables is known and the situation of the species (their presence or their absence) is the response to that gradient.

Hence, the data corresponding to each model located in each site occupy a line in a spreadsheet. The environmental variables, such as rainfall or indoor temperature, are inserted in its columns (Table 3). Last, also in columns, the information corresponding to each model about the presence or absence of the predefined species in each site, or about the presence or absence of the architectonic characteristics in each model and site, as occurs in the present research, is introduced.

Thus, both weather and environmental variables correspond to numerical data, while the morphological characteristics are the equivalent to the values of the species (Table 3). This means that the presence of one of these characteristics implies that number one is the value written in the corresponding cell, while its absence means that number zero is written in it.

The dwellings that were analyzed are represented according to the symbols shown in Table 3. The resulting table can be consulted in the Appendix A (Table A5).

The climate classification that was used is the Köppen scale, established in 1884 by Wladimir Köppen [74] and updated in 1936 [75,76]. The information about wind speed was gathered by means of https://es.windfinder.com [77], a database which presents the values obtained by more than 21,000 weather stations since 1999.



\* ΔTemperature = average outdoor temperature − average indoor temperature (◦C); \*\* ΔHumidity = average outdoor humidity − average indoor humidity (%).

In conclusion, this method allows us to link three concepts: locations, species and variables. The locations are represented by black dots in the graphs. Each one is attached to a letter that indicates the dwelling, plus a number that indicates the archaeological site (Appendix A, Table A1); in total, there are ten black points per dwelling, since each dwelling was located in ten archaeological sites. The species are identified by orange dots joined to their corresponding letter (Table 3). Finally, the variables correspond to the green lines. These vectors mark the zone of the graph where the locations and the species that correspond to the higher values of that specific variable are gathered.

There are three rules that must be followed to read these graphs. First, the links between location, species and variables can be concluded by observing the distance between them. The further a location or a specie is from the vector of a variable, the smaller the influence of that variable is on that location or specie. Second, the closer a location or a specie is to the coordinate origin, the more significant is its presence in the group. Third, the length of a vector depends on the amount of information about its variable that is present in the graph. The longer a vector is, the more information about that variable is contained in the graph.
