750 likes | 1.11k Views
2. Count Data Models. Used for modelling the count of things as a function of covariates. Counts are non-negative integers.. 3. Examples:. Count of vehicles in a queue Number of defective entities Number of failures Number of computers/cars/telephones/etc. in a household. 4. Why special' method
E N D
1. 1 Count Data Models Simon Washington
2. 2 Count Data Models Used for modelling the count of things as a function of covariates. Counts are non-negative integers.
3. 3 Examples: Count of vehicles in a queue
Number of defective entities
Number of failures
Number of computers/cars/telephones/etc. in a household
4. 4 Why ‘special’ methodology? OLS regression can/will predict values that are negative and will also predict non-integer values
There are a number of ways to model counts, but Poisson and negative binomial are ‘popular’.
Also, zero-inflated model can work under certain circumstances
5. 5 Poisson Models
6. 6 Expressions b/w Response and Predictors The expression shown is the log-linear form—there are others as well, but log-linear is most common.
It is the exp portion of the expression that constrains the model forecasts to be positive.
7. 7 Poisson Models
8. 8 Poisson Model Elasticities
9. 9 Why are elasticities useful?i.e., why not just use coefficient estimates?
10. 10 Pseudo Elasticity
11. 11
12. 12 Poisson Models, GOF Measures
13. 13 Poisson Models, GOF Measures
14. 14 Poisson Models, GOF Measures
15. 15
16. 16 Resultant Model
17. 17
18. 18
19. 19 Poisson Model Restriction
20. 20 Over-dispersion
21. 21 Over-dispersion (cntd.)
22. 22 Poisson & Negative Binomial Models: Mathematical Development:
23. 23 Negative Binomial Models:
24. 24 Negative Binomial Model:
25. 25 Test for Overdispersion
26. 26 Test for Overdispersion
27. 27
28. 28
29. 29
30. 30 Censoring & Truncation:
31. 31 Examples
32. 32 Fixed and random effects:
33. 33 fixed and random effects: examples
34. 34 Unobserved Heterogeneity
35. 35 Zero inflated probability (ZIP)
36. 36 Zero inflated probability (ZIP)
37. 37 Poisson and Negative Binomial ZIP Models
38. 38 Vuong’s Statistic
39. 39 Vuong’s Statistic (cntd.)
40. 40 Vuong’s Statistic
41. 41
42. 42
43. 43 Sample Selection Models
44. 44 Count Model: Example 2
45. 45 Count Model Example:
46. 46 Count Model Example:
47. 47
48. 48
49. 49 Count Model Example:
50. 50
51. 51
52. 52
53. 53
54. 54
55. 55 Count Model Example
56. 56
57. 57 Count Model Example:
58. 58 Count Model Example:
59. 59
60. 60
61. 61
62. 62
63. 63
64. 64
65. 65
66. 66 Count Model Example:
67. 67
68. 68 Count Model Example:
69. 69