# Can I use the Estimate of a model directly without considering the p-value?

this is a follow-up question of (this question). The research question is to get the intraindividual (within-subject) variations on contact patterns per period. Each participant reports their contact information for at least two time points. There are three periods, the baseline is before the lockdown.

A plot about the contact rates per age group and location is like this:

Zero contacts are added to the data. When a participant reports contact at the location of the household, it means that there is no contact at the other locations, therefore, the outcome variable (number of contacts) includes a lot of zero.

The model I construct is like this:

glmmTMB(n ~ period * age_4cat * ( sex + I(hh_size) + region +
employment_2cat + education_2cat + weekend + marital_2cat) +
(1 | token) + (1 | Bundesland) + (1 | age_4cat),
family = nbinom2, ziformula = ~1,
# because outcome has high proportion of 0
data = data_education)

# age_4cat is the four age categories
# data_education is the contact information at the location of Education


Then the summary of the model is a long list. This is part of the result that answer the research question.

period2:age_4cat0-25:sexFemale                             -0.093847        NaN     NaN      NaN
period3:age_4cat0-25:sexFemale                              0.033543   0.937228   0.036    0.971
period2:age_4cat26-45:sexFemale                             0.160372   3.484614   0.046    0.963
period3:age_4cat26-45:sexFemale                            -0.178296   1.997667  -0.089    0.929
period2:age_4cat46-65:sexFemale                             0.738602        NaN     NaN      NaN
period3:age_4cat46-65:sexFemale                             0.492707        NaN     NaN      NaN
period2:age_4cat0-25:I(hh_size)                             0.509198        NaN     NaN      NaN
period3:age_4cat0-25:I(hh_size)                            -0.049860   0.280216  -0.178    0.859
period2:age_4cat26-45:I(hh_size)                            0.542733        NaN     NaN      NaN
period3:age_4cat26-45:I(hh_size)                           -0.343575        NaN     NaN      NaN
period2:age_4cat46-65:I(hh_size)                            0.647815   1.139068   0.569    0.570
period3:age_4cat46-65:I(hh_size)                            0.113820        NaN     NaN      NaN
period2:age_4cat0-25:regionEast Germany                    -0.184194   0.177052  -1.040    0.298
period3:age_4cat0-25:regionEast Germany                    -0.358630   2.630182  -0.136    0.892
period2:age_4cat26-45:regionEast Germany                    0.293064        NaN     NaN      NaN
period3:age_4cat26-45:regionEast Germany                    0.730155   2.832746   0.258    0.797
period2:age_4cat46-65:regionEast Germany                   -0.534995        NaN     NaN      NaN
period3:age_4cat46-65:regionEast Germany                    0.141923   2.951278   0.048    0.962
period2:age_4cat0-25:employment_2catNot Employed            0.262844        NaN     NaN      NaN
period3:age_4cat0-25:employment_2catNot Employed            0.059230   2.281125   0.026    0.979
period2:age_4cat26-45:employment_2catNot Employed           0.344246        NaN     NaN      NaN
period3:age_4cat26-45:employment_2catNot Employed           0.258395   1.535815   0.168    0.866
period2:age_4cat46-65:employment_2catNot Employed           0.134121        NaN     NaN      NaN
period3:age_4cat46-65:employment_2catNot Employed          -0.725622        NaN     NaN      NaN
period2:age_4cat0-25:education_2catNon-Higher Education     0.448545        NaN     NaN      NaN
period3:age_4cat0-25:education_2catNon-Higher Education     0.339487        NaN     NaN      NaN
period2:age_4cat26-45:education_2catNon-Higher Education   -0.719402        NaN     NaN      NaN
period3:age_4cat26-45:education_2catNon-Higher Education    0.112603   1.560972   0.072    0.942
period2:age_4cat46-65:education_2catNon-Higher Education    0.291454        NaN     NaN      NaN
period3:age_4cat46-65:education_2catNon-Higher Education   -0.437684   1.626772  -0.269    0.788
period2:age_4cat0-25:weekendNot weekend                     0.085879        NaN     NaN      NaN
period3:age_4cat0-25:weekendNot weekend                     0.366046   3.780621   0.097    0.923
period2:age_4cat26-45:weekendNot weekend                   -0.063371        NaN     NaN      NaN
period3:age_4cat26-45:weekendNot weekend                    0.176045   2.319626   0.076    0.940
period2:age_4cat46-65:weekendNot weekend                   -0.043144   3.167738  -0.014    0.989
period3:age_4cat46-65:weekendNot weekend                    0.438728   2.939998   0.149    0.881
period2:age_4cat0-25:marital_2catmarried/childparticipant   0.043230        NaN     NaN      NaN
period3:age_4cat0-25:marital_2catmarried/childparticipant   0.575552        NaN     NaN      NaN
period2:age_4cat26-45:marital_2catmarried/childparticipant  0.002831   1.667829   0.002    0.999
period3:age_4cat26-45:marital_2catmarried/childparticipant  0.332741   2.255592   0.148    0.883
period2:age_4cat46-65:marital_2catmarried/childparticipant  0.421758   3.567127   0.118    0.906
period3:age_4cat46-65:marital_2catmarried/childparticipant  0.004881   1.756693   0.003    0.998


What I want to get from this result is that at the location of Education, between three periods and four age groups, what are the variables associated with the variations of the number of contacts at the individual level.

The results is strange, there is only one statistically significant variable, and many NaN. Can I use the Estimate to answer the research question?

period2:age_4cat0-25:sexFemale                             -0.093847        NaN     NaN      NaN
period3:age_4cat0-25:sexFemale                              0.033543   0.937228   0.036    0.971


Like for the result above, can I say that the age group 0-25, being female tends to have fewer contacts during period 2 and higher contacts during period 3?

• What do the columns in your output table represent? Point estimates, standard error, and p-value? Commented Aug 1 at 16:01
• @Durden Yes, Estimate, Std. Error, z value, Pr(>|z|)
– Chao
Commented Aug 1 at 17:04
• Why are so many of them undefined? Commented Aug 1 at 18:21
• I think it may be because 80% of the outcome are 0
– Chao
Commented Aug 2 at 7:27