Introduction
In Constructing a ZIP Regression Model, we showed how to use Real Statistics’ Zero-Inflated Poisson Regression tool. We now explain two options: (1) Use of the Init Coeff Range field and (2) LLmin options.
Init Coeff Range
For Example 1, we left the Init Coeff Range blank in Figure 3 Constructing a ZIP Regression Model. This instructed the Zero-Inflated Poisson data analysis tool to use all the regressors for both the Poisson and Logistic models and to set all these coefficients initially to zero.
We could instead fill in this field with a k+1 × 2 range (where k = the # of regressors). This serves two purposes: (1) we can use different regressors for the Poisson and logistic regression models and (2) we can initialize the coefficients with values different from zero.
Solver uses any numeric value in the Init Coeff Range as its initial value and Solver doesn’t use any non-numeric values (i.e. it sets them to zero and doesn’t allow them to change).
Example
Example 1: Rerun Example 1 of Constructing a ZIP Regression Model using only the camper and child regressors for the Poisson regression model and only the persons regressor for the logistic regression model.
We do this by creating a 4 × 2 range as shown in range G2:H5 of Figure 1. We then insert G2:H5 in the Init Coeff Range field of the dialog box that appears in Figure 3 of Constructing a ZIP Regression Model. Note that the row and column labels are not included in this field. Numeric values must be specified for the initial intercept coefficients.
The first part of the output is also shown in Figure 1. We see that only coefficients initialized with numeric values are used in the model. Similarly, the covariance matrix (not displayed) takes the form of a 3+2 × 3+2 range.
Figure 1 – Revised ZIP regression analysis
LLmin Options
As explained in Constructing a ZIP Regression Model, the Real Statistics ZIP Regression data analysis tool provides three options for calculating LLmin. On that webpage we chose the middle option. Figure 2 shows the result if we choose the third option, All Logit, for Example 1 from Constructing a ZIP Regression Model.
Figure 2 – All Logit option
Similarly, for Example 1 above we chose the Intercepts Only option, with the results shown in range N4:O7 of Figure 1. If instead we choose the All Logit option, we obtain the results shown in Figure 3.
Figure 3 – All Logit option for Example 1
Examples Workbook
Click here to download the Excel workbook with the examples described on this webpage.
References
Hilbe, J. M. (2014) Modeling count data. Cambridge University Press
https://assets.cambridge.org/97811070/28333/frontmatter/9781107028333_frontmatter.pdf
Hintze, J. L. (2007) Zero-Inflated Poisson regression. NCSS
https://www.ncss.com/wp-content/themes/ncss/pdf/Procedures/NCSS/Zero-Inflated_Poisson_Regression.pdf
Long, J. S. and Freese, J. (2001) Regression models for categorical dependent variables using Stata
http://investigadores.cide.edu/aparicio/data/refs/Long%26Freese_RegModelsUsingStata_2001.pdf