Here is the data of my regression : 
y is the number of passengers at platform of the train station in each 2 minutes period while A1 to A17 are the number of passengers at 17 study areas on concourse. Time lag has already between considered by shifting the Xs. Since sometimes, there will be no one waiting in the study areas on concourse, so excess zero occurs. I am planing to use zero inflated model. I have tried the code as shown between, but it said "minimum count is not zero" What does that mean and how can i solve it? I have done poisson and it's alright but zero inflated doesn't work.
> setwd('C:/Users/zuzymelody/Desktop')
> try<-read.csv('0inflated_2mins27peak.csv',header=TRUE)
> attach(try)
> names(try)
[1] "y" "A1" "A2" "A3" "A4" "A5" "A6" "A7" "A8" "A9" "A10" "A11"
[13] "A12" "A13" "A14" "A15" "A16" "A17"
> model1<-glm(y~A1+A2+A3+A4+A5+A6+A7+A8+A9+A10+A11+A12+A13+A14+A15+A16+A17,family="poisson")
> summary(model1)
Call:
glm(formula = y ~ A1 + A2 + A3 + A4 + A5 + A6 + A7 + A8 + A9 +
A10 + A11 + A12 + A13 + A14 + A15 + A16 + A17, family = "poisson")
Deviance Residuals:
Min 1Q Median 3Q Max
-7.8598 -3.4571 -0.3663 2.1867 12.5183
Coefficients:
Estimate Std. Error z value Pr(>|z|)
(Intercept) 6.102009 0.164497 37.095 < 2e-16 ***
A1 -0.017555 0.003665 -4.790 1.66e-06 ***
A2 -0.026101 0.017569 -1.486 0.137371
A3 -0.179988 0.014976 -12.018 < 2e-16 ***
A4 -0.032584 0.007735 -4.213 2.52e-05 ***
A5 -0.019908 0.007014 -2.839 0.004532 **
A6 -0.044144 0.010266 -4.300 1.71e-05 ***
A7 0.049829 0.006518 7.645 2.09e-14 ***
A8 -0.080712 0.009819 -8.220 < 2e-16 ***
A9 0.007390 0.007105 1.040 0.298273
A10 0.041116 0.004085 10.065 < 2e-16 ***
A11 -0.041420 0.008418 -4.921 8.62e-07 ***
A12 -0.008241 0.007304 -1.128 0.259171
A13 -0.033161 0.008966 -3.699 0.000217 ***
A14 0.020818 0.005250 3.965 7.34e-05 ***
A15 -0.002995 0.006125 -0.489 0.624887
A16 -0.061997 0.017122 -3.621 0.000294 ***
A17 -0.025025 0.008391 -2.982 0.002860 **
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
(Dispersion parameter for poisson family taken to be 1)
Null deviance: 1137.71 on 29 degrees of freedom
Residual deviance: 599.74 on 12 degrees of freedom
AIC: 840.1
Number of Fisher Scoring iterations: 5
>with(model1, cbind(res.deviance = deviance, df = df.residual,
p = pchisq(deviance, df.residual, lower.tail=FALSE)))
res.deviance df p
[1,] 599.7445 12 1.202013e-120
> require( pscl )
> Zip<-zeroinfl(model1,link="logit",dist="poisson")
**Error in zeroinfl(model1, link = "logit", dist = "poisson") :
invalid dependent variable, minimum count is not zero**
dput(try) structure(list(y = c(156L, 74L, 221L, 207L, 168L, 36L, 128L, 208L, 99L, 117L, 228L, 211L, 341L, 173L, 196L, 310L, 112L, 203L, 104L, 183L, 325L, 143L, 218L, 166L, 218L, 127L, 136L, 38L, 102L, 34L), A1 = c(24L, 24L, 24L, 19L, 20L, 9L, 14L, 23L, 15L, 23L, 14L, 16L, 15L, 25L, 25L, 19L, 24L, 26L, 25L, 26L, 22L, 14L, 13L, 15L, 9L, 12L, 9L, 12L, 15L, 18L), A2 = c(2L, 4L, 0L, 3L, 0L, 1L, 1L, 2L, 1L, 2L, 0L, 2L, 2L, 0L, 1L, 1L, 3L, 3L, 2L, 2L, 3L, 2L, 3L, 5L, 4L, 3L, 4L, 1L, 2L, 1L), A3 = c(2L, 2L, 0L, 1L, 1L, 9L, 3L, 0L, 0L, 0L, 1L, 1L, 3L, 1L, 0L, 0L, 1L, 2L, 3L, 1L, 0L, 1L, 1L, 0L, 1L, 1L, 1L, 0L, 0L, 2L), A4 = c(15L, 11L, 6L, 7L, 10L, 10L, 5L, 4L, 5L, 7L, 9L, 9L, 4L, 6L, 6L, 13L, 9L, 13L, 9L, 10L, 6L, 6L, 7L, 6L, 10L, 9L, 10L, 7L, 9L, 2L), A5 = c(13L, 10L, 6L, 6L, 11L, 19L, 13L, 14L, 7L, 7L, 6L, 8L, 10L, 5L, 7L, 9L, 9L, 11L, 3L, 13L, 8L, 8L, 8L, 6L, 8L, 9L, 9L, 14L, 9L, 6L), A6 = c(9L, 10L, 9L, 9L, 4L, 7L, 7L, 12L, 11L, 11L, 12L, 8L, 6L, 7L, 8L, 5L, 9L, 6L, 5L, 6L, 9L, 11L, 6L, 6L, 8L, 9L, 4L, 11L, 10L, 7L ), A7 = c(21L, 16L, 13L, 13L, 4L, 9L, 12L, 13L, 12L, 12L, 12L, 6L, 7L, 6L, 6L, 4L, 5L, 9L, 8L, 7L, 9L, 12L, 10L, 7L, 8L, 12L, 14L, 2L, 6L, 6L), A8 = c(1L, 5L, 10L, 10L, 1L, 9L, 6L, 6L, 7L, 7L, 5L, 6L, 3L, 2L, 4L, 0L, 4L, 2L, 5L, 5L, 5L, 3L, 2L, 4L, 3L, 8L, 10L, 8L, 2L, 5L), A9 = c(8L, 9L, 10L, 10L, 12L, 19L, 10L, 6L, 6L, 6L, 0L, 6L, 8L, 10L, 2L, 3L, 6L, 2L, 2L, 6L, 5L, 2L, 4L, 1L, 3L, 7L, 7L, 4L, 4L, 2L), A10 = c(7L, 10L, 12L, 20L, 24L, 21L, 24L, 18L, 20L, 18L, 26L, 21L, 12L, 11L, 18L, 18L, 19L, 16L, 25L, 21L, 22L, 14L, 12L, 17L, 21L, 14L, 14L, 10L, 8L, 7L), A11 = c(0L, 2L, 1L, 4L, 2L, 1L, 1L, 1L, 13L, 10L, 12L, 5L, 2L, 0L, 5L, 1L, 4L, 4L, 3L, 3L, 1L, 1L, 3L, 3L, 5L, 5L, 2L, 10L, 3L, 4L), A12 = c(12L, 14L, 14L, 17L, 10L, 14L, 13L, 19L, 7L, 5L, 6L, 6L, 8L, 7L, 13L, 11L, 10L, 8L, 6L, 6L, 9L, 14L, 9L, 10L, 8L, 9L, 8L, 9L, 5L, 7L ), A13 = c(6L, 2L, 1L, 5L, 9L, 6L, 7L, 4L, 12L, 5L, 9L, 10L, 3L, 7L, 4L, 2L, 2L, 6L, 4L, 6L, 7L, 4L, 9L, 6L, 11L, 4L, 5L, 4L, 6L, 6L), A14 = c(14L, 13L, 16L, 11L, 8L, 6L, 9L, 13L, 14L, 14L, 9L, 8L, 12L, 11L, 13L, 11L, 18L, 15L, 20L, 21L, 17L, 18L, 18L, 18L, 25L, 20L, 12L, 9L, 8L, 8L), A15 = c(7L, 6L, 7L, 5L, 4L, 9L, 12L, 12L, 11L, 12L, 9L, 8L, 7L, 8L, 10L, 16L, 8L, 8L, 13L, 10L, 5L, 5L, 8L, 10L, 10L, 4L, 6L, 6L, 6L, 7L), A16 = c(2L, 1L, 3L, 3L, 1L, 2L, 3L, 2L, 3L, 2L, 2L, 1L, 2L, 2L, 3L, 3L, 2L, 1L, 3L, 4L, 2L, 5L, 4L, 8L, 5L, 2L, 1L, 2L, 2L, 2L), A17 = c(10L, 13L, 13L, 2L, 5L, 1L, 3L, 3L, 5L, 4L, 4L, 6L, 4L, 6L, 3L, 2L, 2L, 2L, 7L, 8L, 3L, 7L, 5L, 6L, 7L, 6L, 6L, 3L, 4L, 3L)), .Names = c("y", "A1", "A2", "A3", "A4", "A5", "A6", "A7", "A8", "A9", "A10", "A11", "A12", "A13", "A14", "A15", "A16", "A17"), class = "data.frame", row.names = c(NA, -30L))
above is the reproducible example. Sorry its my first time to post here, dont know the rule well
var(y) >> mean(y), which is the $\lambda$ parameter of the Poisson distribution. For a Poisson distribution to be effectvar(y) = mean(y). – akash87