Skip to main content

Table 3 Comparison of independent mixture models fitted to parasite and leukocyte counts by AIC and BIC

From: Evidence for overdispersion in the distribution of malaria parasites and leukocytes in thick blood smears

 

Poisson mixture

Negative binomial mixture

m =1

-â„’

AIC

BIC

-â„’

AIC

BIC

p a

6801.59

13605.17

13609.80

3200.63

6405.25

6414.50

p b

10838.95

21679.91

21684.75

4344.27

8692.54

8702.23

p c

2472.18

4946.36

4951.08

2302.96

4609.92

4619.38

â„“ a

3108.25

6218.51

6223.13

2532.77

5069.53

5078.79

â„“ b

3547.53

7097.06

7101.90

2965.34

5934.69

5944.38

â„“ c

3051.08

6104.15

6108.88

2728.46

5460.91

5470.37

m =2

-â„’

AIC

BIC

-â„’

AIC

BIC

p a

3962.18

7930.35

7944.23

3200.63

6409.25

6430.53

p b

5882.41

11770.81

11785.34

4344.27

8696.54

8718.69

p c

2289.73

4585.47

4599.65

2302.96

4613.93

4635.61

â„“ a

2633.87

5273.75

5287.62

2532.77

5073.54

5094.81

â„“ b

3029.67

6065.33

6079.86

2965.35

5938.69

5960.84

â„“ c

2756.98

5519.97

5534.15

2728.45

5464.91

5486.59

m =3

-â„’

AIC

BIC

-â„’

AIC

BIC

p a

3397.75

6805.50

6828.63

3200.63

6413.25

6447.60

p b

4761.19

9532.38

9556.60

4344.27

8700.54

8736.20

p c

2288.39

4586.77

4610.41

2302.96

4617.93

4652.89

â„“ a

2527.85

5065.70

5088.83

2532.77

5077.54

5111.88

â„“ b

2945.87

5901.74

5925.95

2965.35

5942.69

5978.35

â„“ c

2729.21

5468.42

5492.06

2728.45

5468.90

5503.87

m =4

-â„’

AIC

BIC

-â„’

AIC

BIC

p a

3267.46

6548.92

6581.29

3189.16

6394.32

6442.42

p b

4470.16

8954.33

8988.24

4344.27

8704.54

8754.38

p c

2288.21

4590.42

4623.52

2302.96

4621.93

4670.85

â„“ a

2519.22

5052.44

5084.81

2532.77

5081.54

5129.63

â„“ b

2938.52

5891.05

5924.95

2965.35

5946.69

5996.53

â„“ c

2721.23

5456.47

5489.57

2728.45

5472.90

5521.82

  1. Parasite (p a , p b , p c ) and leukocyte ( â„“ a , â„“ b , â„“ c ) counts are fitted to Poisson mixtures and negative binomial mixtures. The number of components is m . Minus log-likelihood (-â„’) and information measures (AIC and BIC) are given. Models were fitted by maximum likelihood using the expectation-maximization (EM) algorithm, and validated by direct numerical maximization using nlm in R.