xtpoisson interpretation

[Rebecca England]

Hi there,

I am using a poisson regression to regress longitudinal data, I have a count variable for my DV (ypnpal - number of close friends) and I am regressing sports participation intensity data on it (individual vs team sports x time spent doing sport each week).

Am I correct in thinking that I can interpret the coefficients as though my DV is log(ypnpal) = X... ?

My regressions also give no constant value, but I haven't suppressed this, any ideas why it wouldn't give this output?

I'll provide an example of one of my regressions below. For context I include separate individual and team sport variables as these are not mutually exclusive binary variables.

. xtpoisson ypnpal teamsport_time indivsport_time, fe
note: 5614 groups (5614 obs) dropped because of only one obs per group
note: 22 groups (47 obs) dropped because of all zero outcomes

Iteration 0: log likelihood = -42471.606
Iteration 1: log likelihood = -42356.056
Iteration 2: log likelihood = -42356.055

Conditional fixed-effects Poisson regression Number of obs = 18,996
Group variable: pidp Number of groups = 7,520

Obs per group:
min = 2
avg = 2.5
max = 4

Wald chi2(2) = 230.61
Log likelihood = -42356.055 Prob > chi2 = 0.0000

---------------------------------------------------------------------------------
ypnpal | Coefficient Std. err. z P>|z| [95% conf. interval]
----------------+----------------------------------------------------------------
teamsport_time | .0187247 .0027242 6.87 0.000 .0133853 .024064
indivsport_time | -.038564 .0025406 -15.18 0.000 -.0435435 -.0335845
---------------------------------------------------------------------------------

Other regressions include different lags for each variable. But poisson is new to me and I'm not sure how to interpret things like the "prob > chi2".

[Fei Wang]

Am I correct in thinking that I can interpret the coefficients as though my DV is log(ypnpal) = X... ?

Roughly speaking, yes. The coefficients are semi-elasticities.

My regressions also give no constant value, but I haven't suppressed this, any ideas why it wouldn't give this output?

FE poisson doesn't report an overall constant. You may treat the FEs as the subject-specific constant.

I'm not sure how to interpret things like the "prob > chi2".

It's a joint significance test for the coefficients of all X's. "Prob > chi2" reports the p value for the test.