At the beginning of the blog, you wrote:

I actually prefer restricted cubic splines over polynomials, so it wouldn’t bother me if people did these as a default replacement for the polynomial functions. Also see this paper by Francis et al. using b-splines.

What the difference between restricted cubic splines and b-splines? Can you recommend any references introducing b-splines.

I am new to spline modelling.

Thanks!

Pauline

]]>Thank you so much. Works great!

]]>The spreadsheet approach is maybe not so good if you have many years. Here is a simple python script, it will not be uber efficient, but will still be OK even if you have millions of years and many records.

#################################

import numpy as np

# arrays that have begin and end years

begin = np.array([2005,2010,2016])

end = np.array([2005,2011,2020])

toty = end – begin + 1 # assumes no partial years

# now make arrays to hold the aoristic results

low = begin.min()

hig = end.max()

yr = np.arange(low,hig+1)

res = []

for y in yr:

# See how many are inside years

inside = (y >= begin) & (y <= end)

# weight cases by toty

av = (inside/toty).sum()

res.append(av)

res = np.array(res)

# Show the results, these will sum to the original counts

np.stack([yr,res],axis=1)

#################################

I am an archaeology student from Switzerland and was wondering whether your spreadsheet can be rewritten, so that the timespan of BeginOrig and EndOrig would cover Years instead of Days/Hours or Weeks.

Cheers ]]>

Very nice, thank you Justin. I was surprised that SUNY was 50% — UNO is even higher!

Good idea with the map as well, I bet many grad students don’t realize you need to be willing to move to wherever. (Although if folks want to stay, Omaha I believe will grow a decent bit in private sector in say the next decade I bet, several big insurance companies centered there.)

]]>