Best way to affecting the associated match weights for different comparisons?

[mmagoffin-sd]

I've been successful in outputting linkage results for matching records, but I was wondering what the best way to affect how much a column is weighted would be? For example, looking through my possible matches from linker.predict(), I see that it looks like last names not matching is not pulling the overall match weight score down enough to make a match as unlikely as I would reasonably expect. Would a solution to this lie in manipulating the blocking rules or comparison levels somehow?

[RobinL]

You can manually set the m and u probabilities in your settings object if you want to override the estimated values.

For example you can see in the last cell here we're saving settings to json.

And then loading in the first cell here

And in the saved settings you can see the m and u probabilities. These are the values you'd want to change for a manual override.

More information here on what the m and u probabilities mean

You can also manually set the m and u probabilities with code like:

# Get first comparison
comp = linker._settings_obj.comparisons[0]

# get first comparison level
cl = comp.comparison_levels[1]

# set m probability
cl.m_probability = 0.1