Cheat Sheet: Interpreting Regressions: L (P (Y X X
Cheat Sheet: Interpreting Regressions: L (P (Y X X
OLS None Continuous Continuous Y = α+ βx + ε β: the effect of 1 unit Δ in x on Δ in y ◦ Basic regression to estimate relationships
OLS Log-‐Linear Continuous Continuous ln(Y) = α+ βx + ε β: the effect of 1 unit Δ in x on %Δ in y ◦ Easier interpretation, non-‐linear
OLS Linear-‐Log Continuous Continuous Y = α+ βln(x) + ε β: the effect of 1 % Δ in x on Δ y ◦ Easier interpretation, non-‐linear
OLS Log-‐Log Continuous Continuous ln(Y) = α+ βln(x) + ε β: the effect of 1 % Δ in x on %Δ in y ◦ Interpret as an elasticity, non-‐linear
Dummy
OLS None Continuous Y
=
α+
βx
+
ε
β:
the
effect
of
x
being
equal
to
1
on
Δ
in
y ◦
Think
of
X
as
switched
on
and
off
(X
=
0
or
1)
Dummy
Dummy β:
the
effect
of
x
being
equal
to
1
on
probability
OLS None Y
=
α+
βx
+
ε
◦
Think
of
X
and
Y
as
switched
on
and
off
(X
=
0
or
1) (Y
=
0
or
1)
that
Y
occurs
(a
%)
β1:
the
effect
of
a
1
unit
Δ
in
x1
on
Δ
in
y,
controlling
for
the
effect
of
x2 ◦
Control
for
variables
Any Multivariate Continuous Continuous Y
=
α+
β1x1
+
β2x2
+
ε
◦
Incorporate
effect
of
multiple
variables
β2:
the
effect
of
a
1
unit
Δ
in
x2
on
Δ
in
y,
controlling
for
the
effect
of
x1
Y
=
α+
βZ
+
ε
Continuous
or
Continuous
2SLS Any where
Corr(Z,
X)>0
β:
t
he
c
ausal
e
ffect
of
1
unit
Δ
in
x
on
Δ
in
y ◦
Causal
interpretation
of
our
β
coefficient
Dummy or
Dummy
and
Corr(Z,
ε)
=
0