more host data: length of stay and rating
some musette de cour. a lovely painting of a bagpiper too
i won’t do much discussion this time. this is a short test i did to the data and it somewhat confirms the rule that some hosts have that there should be a minimum stay. this is the first widespread rule that i see has some evidence to it (even though the correlation is very low, 0.17454, octave source)
here is some evidence that rating is biased: the longer someone stays, the more likely it is that their rating will be higher. it doesn’t mean, however, that the maximum rating will be affected. instead, it seems there is a “logarithmic” effect on the curves: though it grows, it will settle on a value that still might not reach a high rating, and probably settles on what would be the “final evaluation” of the experience. i.e., you don’t like someone at first sight, but with time, you will gain a better perspective. this perspective might still be negative or positive, so this is not a predictor of quality of stay. instead, it is probably a predictor of how good it will be seen in the end.
for this set all guests were considered, including friends of CSers and non-CS people.
an interesting test of this would be a rating of a guest over time. this would give us an idea of how personal rating changes over time. i have some data on another rating i did a while a go, i might stitch it together with this one. here is the plot of the data
still fuzzy, but a much nicer shape.