Citazione Originariamente Scritto da Borat Visualizza Messaggio
The strongest evidences in that paper are table 3 and the regression.

I have criticized the regression because it does not use the available rural data.

Computing

(tmin_u - tmin_r) = a + b*pop^(1/4)

would be easy and would purify the results from variations in free atmosphere and environment temperature.

Perhaps they had not as many year of rural data as a good regression requires (in political science analyses of 23 cases are often published also in top journals, but this is much sub-optimal: one should not rely on asymptotic properties of OLS estimators below 50 observations and personally I avoid falling below 2-300 observations).

Table 3 is this
Immagine


What is questionable in this table is that the stations are at different altitudes, with differences up to 100-130 meters that can account for about 1°C if we assume the typical vertical temperature gradient of 0.65°C/100 m.

This evening if I haven't anything better to do I'll compute a rescaled version of that table, where the effects of differential altitudes are purified (in practice, I add -0.65°C for every 100 meters difference altitude to all entries). The uhi depicted by the table will be less drastic.

With me as a referee, that article wouldn't get published. Problems in regression, no supply of information with reasonable correction for altitude etc etc.

Agreed because the altitude is a substantial issue. But how do we sovle the case of direct comparility of the whole Thiseio dataset with a rural area?

Note than in 1860 Athens was not more than 50.000(?) residents I think....