Skip to content

linearmodels/weights-column-name

Rule weights-column-name
Pack linearmodels
Severity error
Fix unsafe, with xpd check --fix --unsafe-fixes
Library versions every version

weights takes the values and not the name of a column, so a name raises. Pass the column itself, as in weights=df["w"].

model = PooledOLS.from_formula("lwage ~ 1 + exper", data, weights="hours")
model = PooledOLS.from_formula("lwage ~ 1 + exper", data, weights=data["hours"])

The rule reports weights="name" on a model or on its from_formula. statsmodels and R take the weights of a formula model as the name of a column, but linearmodels takes the values.

xpd check --fix --unsafe-fixes applies the fixes of this rule. Check the result, because an unsafe fix can change what the code does.

The unsafe fix writes weights=data["hours"] for weights="hours" in from_formula when the data is a name. The call raised before the fix. A model built from arrays has no fix, because the frame is not in the call.

7.1 raised TypeError: weights must be array_like ... Got str on PooledOLS and on IV2SLS, and 7.0 raised AttributeError: 'str' object has no attribute 'ndim'.

xpd prints this message with the issue. A name that starts with $ stands for code from your file.

weights takes the values, not the name of a column, and $W raises TypeError on linearmodels 7.1 and AttributeError on 7.0. Write weights=$D[$W], the column itself.

Browse the other rules of the linearmodels pack.