Skip to main content
. Author manuscript; available in PMC: 2017 Sep 29.
Published in final edited form as: Am Sociol Rev. 2017 May 31;82(3):568–599. doi: 10.1177/0003122417706391

Table 4.

Log-Linear Model Fit Statistics

Model df L2 p-value BIC Δ
Men (N = 56,200)
Baseline Models
0a: Origin + Destination + Period 27,972 244,675 <.01 −61,245 69.0
1a: 0a + Origin×Period + Destination×Period 27,380 239,997 <.01 −59,449 68.5
2a: 1a + Parameter-ABCD 27,292 27,431 .28 −271,052 18.2
Period-Varying Topological Mobility Models
3a: 2a + Period×Parameter-A 26,992 24,874 >.99 −270,328 14.2
4a: 2a + Period×Parameter-AB 26,956 24,723 >.99 −270,086 14.1
5a: 2a + Period×Parameter-ABC 26,944 24,661 >.99 −270,017 14.1
6a: 2a + Period×Parameter-ABCD 26,940 24,648 >.99 −269,986 14.1
7a: 2a + Period×Parameter-A'B'C'D' 27,276 25,659 >.99 −272,650 16.0
Women (N = 47,180)
Baseline Models
0b: Origin + Destination + Period 27,972 174,593 <.01 −126,434 66.6
1b: 0b + Origin×Period + Destination×Period 27,380 166,191 <.01 −128,465 65.7
2b: 1b + Parameter-ABCD 27,292 19,635 >.99 −274,074 15.2
Period-Varying Topological Mobility Models
3b: 2b + Period×Parameter-A 26,992 18,287 >.99 −272,193 12.3
4b: 2b + Period×Parameter-AB 26,956 18,154 >.99 −271,939 12.2
5b: 2b + Period×Parameter-ABC 26,944 18,102 >.99 −271,862 12.1
6b: 2b + Period×Parameter-ABCD 26,940 18,077 >.99 −271,844 12.1
7b: 2b + Period×Parameter-A'B'C'D' 27,276 18,879 >.99 −274,657 13.9

Data source: Panel Study of Income Dynamics, 1969 to 2011.

Note: L2 is the log-likelihood ratio chi-square statistic with the degrees of freedom (df) reported in column df. The Bayesian Information Criteria (BIC) = L2 − (df)×log(N), where N is the total number of observations. A: micro-class immobility parameters; B: net micro-class mobility within meso-class parameters; C: net meso-class mobility within macro-class parameters; D: net macro-class mobility within manual/nonmanual sector parameters. Period×Parameter-: interaction between period and mobility parameters. A', B', C', and D' indicate that trend parameters are constrained to be equal across classes at the micro-class, meso-class, macro-class, and sectoral levels, respectively.