1

I have the following logistic regression in R:

glm(FundingSuccess ~ FundingGoal+ VentureTeamDiversity, 
      data = df, family = "binomial")

The VentureTeamDiversity variable is a Blau diversity index that ranges from 0 to 1. The regression coefficient of VentureTeamDiversity is 4.75 (p < 0.01). How do I interpret this coefficient? Basically, I am looking at understanding how much does an increase in VentureTeamDiversity by 10 percent affect the probability of successful funding (holding FundingGoal constant).

kjetil b halvorsen
  • 63,378
  • 26
  • 142
  • 467
SanMelkote
  • 621
  • 5
  • 20

1 Answers1

2

The coefficient $\beta_1$ can be interpreted as follows: $e^{\beta_1t}$ is the ratio of the odds of predicting a 1 after increasing $X_1$ by $t$ units to the odds of predicting a 1 while leaving $X_1$ alone.

In your specific scenario, $e^{4.75(0.10)} = 1.049$ means that the odds of getting funding after a 10% increase in VentureTeamDiversity is $1.049$ times the odds of getting funding without that increase.

There's also this answer How to interpret logistic regression coefficient, which describes how to compute actual changes in the probability of predicting a 1, but that can only be computed at specific values of the predictor.

Nick Koprowicz
  • 838
  • 3
  • 12
  • You could take the individual marginal effects at the observed values of $x$ and take the average over the estimation sample. This would produce a single number that gives you the average change in the probability of success associated with one additional unit of $x$ for everyone in the sample. – dimitriy Dec 10 '21 at 01:29