Calculate odds ratio from linear regression coefficient with binary dependt variable?

[Aleksander Erichsen]

Hello everyone!

I have a multiple linear regression model with, among others, a categorical variable for gender (male / female) which I use to predict the probability of the respondent having submitted an idea (yes/no).
Female is used as the baseline category. Is there a way to use the coeffienct from men to calculate the odds ration between men and women?

I assume that it should be possible since the DV is dichotomous and hence the coefficient is in effect a measure of probability? If so, how would the procedure to calculate this be?

The coefficient for men are: .2485926

Thanks for all the fantastic help provided in this forum!

[Ken Chui]

Is there a reason for not using logistic regression? Like:

logistic IDEA i.MALE X2 X3 X4

would provide the odds ratio directly in the output.

[Joseph Coveney]

I suppose that there's a reason for your fitting a linear probability model, but I agree with Ken that if you're interested in odds ratios, then -logit- seems the more straightforward approach.

Anyway, you could consider something like the following.

.ÿ
.ÿversionÿ17.0

.ÿ
.ÿclearÿ*

.ÿ
.ÿsetÿseedÿ`=strreverse("1606170")'

.ÿ
.ÿquietlyÿsetÿobsÿ250

.ÿ
.ÿgenerateÿbyteÿoutÿ=ÿruniform()ÿ<ÿ0.5

.ÿgenerateÿbyteÿsexÿ=ÿmod(_n,ÿ2)

.ÿgenerateÿdoubleÿpreÿ=ÿruniform(-0.5,ÿ0.5)

.ÿ
.ÿ*
.ÿ*ÿBeginÿhere
.ÿ*
.ÿregressÿoutÿi.sexÿc.preÿ//ÿ,ÿvce(robust)

ÿÿÿÿÿÿSourceÿ|ÿÿÿÿÿÿÿSSÿÿÿÿÿÿÿÿÿÿÿdfÿÿÿÿÿÿÿMSÿÿÿÿÿÿNumberÿofÿobsÿÿÿ=ÿÿÿÿÿÿÿ250
-------------+----------------------------------ÿÿÿF(2,ÿ247)ÿÿÿÿÿÿÿ=ÿÿÿÿÿÿ0.48
ÿÿÿÿÿÿÿModelÿ|ÿÿÿ.24304012ÿÿÿÿÿÿÿÿÿ2ÿÿÿ.12152006ÿÿÿProbÿ>ÿFÿÿÿÿÿÿÿÿ=ÿÿÿÿ0.6168
ÿÿÿÿResidualÿ|ÿÿ62.0009599ÿÿÿÿÿÿÿ247ÿÿ.251016032ÿÿÿR-squaredÿÿÿÿÿÿÿ=ÿÿÿÿ0.0039
-------------+----------------------------------ÿÿÿAdjÿR-squaredÿÿÿ=ÿÿÿ-0.0042
ÿÿÿÿÿÿÿTotalÿ|ÿÿÿÿÿÿ62.244ÿÿÿÿÿÿÿ249ÿÿ.249975904ÿÿÿRootÿMSEÿÿÿÿÿÿÿÿ=ÿÿÿÿ.50102

------------------------------------------------------------------------------
ÿÿÿÿÿÿÿÿÿoutÿ|ÿCoefficientÿÿStd.ÿerr.ÿÿÿÿÿÿtÿÿÿÿP>|t|ÿÿÿÿÿ[95%ÿconf.ÿinterval]
-------------+----------------------------------------------------------------
ÿÿÿÿÿÿÿ1.sexÿ|ÿÿ-.0572617ÿÿÿ.0634409ÿÿÿÿ-0.90ÿÿÿ0.368ÿÿÿÿ-.1822159ÿÿÿÿ.0676925
ÿÿÿÿÿÿÿÿÿpreÿ|ÿÿÿ-.049069ÿÿÿ.1133506ÿÿÿÿ-0.43ÿÿÿ0.665ÿÿÿÿÿ-.272326ÿÿÿÿÿ.174188
ÿÿÿÿÿÿÿ_consÿ|ÿÿÿ.5609682ÿÿÿ.0448679ÿÿÿÿ12.50ÿÿÿ0.000ÿÿÿÿÿ.4725956ÿÿÿÿ.6493407
------------------------------------------------------------------------------

.ÿmarginsÿsex,ÿpost

PredictiveÿmarginsÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿNumberÿofÿobsÿ=ÿ250
ModelÿVCE:ÿOLS

Expression:ÿLinearÿprediction,ÿpredict()

------------------------------------------------------------------------------
ÿÿÿÿÿÿÿÿÿÿÿÿÿ|ÿÿÿÿÿÿÿÿÿÿÿÿDelta-method
ÿÿÿÿÿÿÿÿÿÿÿÿÿ|ÿÿÿÿÿMarginÿÿÿstd.ÿerr.ÿÿÿÿÿÿtÿÿÿÿP>|t|ÿÿÿÿÿ[95%ÿconf.ÿinterval]
-------------+----------------------------------------------------------------
ÿÿÿÿÿÿÿÿÿsexÿ|
ÿÿÿÿÿÿÿÿÿÿ0ÿÿ|ÿÿÿ.5606309ÿÿÿ.0448358ÿÿÿÿ12.50ÿÿÿ0.000ÿÿÿÿÿ.4723215ÿÿÿÿ.6489402
ÿÿÿÿÿÿÿÿÿÿ1ÿÿ|ÿÿÿ.5033691ÿÿÿ.0448358ÿÿÿÿ11.23ÿÿÿ0.000ÿÿÿÿÿ.4150598ÿÿÿÿ.5916785
------------------------------------------------------------------------------

.ÿnlcomÿOR:ÿexp(logit(_b[1.sex])ÿ-ÿlogit(_b[0.sex]))

ÿÿÿÿÿÿÿÿÿÿOR:ÿexp(logit(_b[1.sex])ÿ-ÿlogit(_b[0.sex]))

------------------------------------------------------------------------------
ÿÿÿÿÿÿÿÿÿÿÿÿÿ|ÿCoefficientÿÿStd.ÿerr.ÿÿÿÿÿÿzÿÿÿÿP>|z|ÿÿÿÿÿ[95%ÿconf.ÿinterval]
-------------+----------------------------------------------------------------
ÿÿÿÿÿÿÿÿÿÿORÿ|ÿÿÿ.7943382ÿÿÿ.2030884ÿÿÿÿÿ3.91ÿÿÿ0.000ÿÿÿÿÿ.3962921ÿÿÿÿ1.192384
------------------------------------------------------------------------------

.ÿ//ÿor
.ÿnlcomÿOR:ÿ(ÿ_b[1.sex]ÿ/ÿ(1ÿ-ÿ_b[1.sex])ÿ)ÿ/ÿ(ÿ_b[0.sex]ÿ/ÿ(1ÿ-ÿ_b[0.sex])ÿ)

ÿÿÿÿÿÿÿÿÿÿOR:ÿ(ÿ_b[1.sex]ÿ/ÿ(1ÿ-ÿ_b[1.sex])ÿ)ÿ/ÿ(ÿ_b[0.sex]ÿ/ÿ(1ÿ-ÿ_b[0.sex])ÿ)

------------------------------------------------------------------------------
ÿÿÿÿÿÿÿÿÿÿÿÿÿ|ÿCoefficientÿÿStd.ÿerr.ÿÿÿÿÿÿzÿÿÿÿP>|z|ÿÿÿÿÿ[95%ÿconf.ÿinterval]
-------------+----------------------------------------------------------------
ÿÿÿÿÿÿÿÿÿÿORÿ|ÿÿÿ.7943382ÿÿÿ.2030884ÿÿÿÿÿ3.91ÿÿÿ0.000ÿÿÿÿÿ.3962921ÿÿÿÿ1.192384
------------------------------------------------------------------------------

.ÿ
.ÿlogitÿoutÿi.sexÿc.pre,ÿorÿnolog

LogisticÿregressionÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿNumberÿofÿobsÿ=ÿÿÿÿ250
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿLRÿchi2(2)ÿÿÿÿ=ÿÿÿ0.98
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿProbÿ>ÿchi2ÿÿÿ=ÿ0.6134
Logÿlikelihoodÿ=ÿ-172.28576ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿPseudoÿR2ÿÿÿÿÿ=ÿ0.0028

------------------------------------------------------------------------------
ÿÿÿÿÿÿÿÿÿoutÿ|ÿOddsÿratioÿÿÿStd.ÿerr.ÿÿÿÿÿÿzÿÿÿÿP>|z|ÿÿÿÿÿ[95%ÿconf.ÿinterval]
-------------+----------------------------------------------------------------
ÿÿÿÿÿÿÿ1.sexÿ|ÿÿÿ.7941829ÿÿÿ.2019731ÿÿÿÿ-0.91ÿÿÿ0.365ÿÿÿÿÿ.4824437ÿÿÿÿ1.307358
ÿÿÿÿÿÿÿÿÿpreÿ|ÿÿÿ.8203934ÿÿÿ.3730382ÿÿÿÿ-0.44ÿÿÿ0.663ÿÿÿÿÿ.3364924ÿÿÿÿÿ2.00018
ÿÿÿÿÿÿÿ_consÿ|ÿÿÿ1.277965ÿÿÿ.2307145ÿÿÿÿÿ1.36ÿÿÿ0.174ÿÿÿÿÿ.8971177ÿÿÿÿ1.820492
------------------------------------------------------------------------------
Note:ÿ_consÿestimatesÿbaselineÿodds.

.ÿ
.ÿexit

endÿofÿdo-file


.

I trust that your predictions aren't beyond the pale.