Common use of Dependent Variables Clause in Contracts

Dependent Variables. Using the 2007-cohort and the RISc-data, we examine whether or not an offender is imprisoned (in/out) and if so, for how long. This approach is consistent with prior research.4 Incarceration is measured with a dichotomous variable coded 0 for non- incarceration and 1 for incarceration, and the length of the prison sentence is a con- tinuous measure capturing the total days of confinement ordered by the judge.5 The in/out decision is not analyzed using the Prison Project data because the vast majority of the suspects in this sample received a prison sentence (94%) (see also ▇▇▇▇▇▇▇ et al., 2010; Ulmer, Eisenstein, and ▇▇▇▇▇▇▇, 2010). Instead, we examine whether or not the unsuspended length of the imposed prison sentence exceeds the time served in pretrial detention.6 For the subgroup of offenders who receive “extra time,” we examine the length of this extra term of imprisonment, and for the full sample, the full length of the prison sentence is measured. Among the total 2007-cohort, 18 percent are incarcerated, with a mean term of imprisonment of 226 days. A prison sentence is imposed in 47 percent of the cases in the RISc-data, and the average prison term is 345 days. Of all offenders in the Prison Project, 61 percent receives a prison length at sentencing that exceeds the term of pretrial detention, with an average “extra term” of 418 days. The full prison length is approximately 370 days. Differences in the above described results are caused by differences in population and measurement between datasets. Important differences are related to the extent through which defendants have been selected through the criminal justice system: the 2007-cohort includes all suspects, the RISc-data only those with a pre-sentencing report, and the Prison Project includes pretrial detained offenders only. Consequently, when the results of ethnic disparities in sentencing are presented, it is especially valuable to pay attention to differences within datasets rather than between. The dichotomous sentencing outcomes are modeled with logistic regression. For those incarcerated, sentence length outcomes are transformed logarithmically and modeled with OLS regression. The log transformation normalizes the skewed distri- bution and addresses the fact that additional days of incarceration may have different meaning for different sentence lengths. Also, it allows for more convenient interpreta- tion of the sentence lengths in terms of their proportional increase associated with a unit increase in each explanatory variable. The regression coefficients are exponenti- ated to provide for proportional increases in the logged sentence length models.

Dependent Variables. Using the 2007-cohort and the RISc-data, we examine whether or not an offender is imprisoned (in/out) and if so, for how long. This approach is consistent with prior research.4 Incarceration is measured with a dichotomous variable coded 0 for non- incarceration and 1 for incarceration, and the length of the prison sentence is a con- tinuous measure capturing the total days of confinement ordered by the judge.5 The in/out decision is not analyzed using the Prison Project data because the vast majority of the suspects in this sample received a prison sentence (94%) (see also ▇▇▇▇▇▇▇ et al., 2010; Ulmer▇▇▇▇▇, Eisenstein▇▇▇▇▇▇▇▇▇▇, and ▇▇▇▇▇▇▇, 2010). Instead, we examine whether or not the unsuspended length of the imposed prison sentence exceeds the time served in pretrial detention.6 For the subgroup of offenders who receive “extra time,” we examine the length of this extra term of imprisonment, and for the full sample, the full length of the prison sentence is measured. Among the total 2007-cohort, 18 percent are incarcerated, with a mean term of imprisonment of 226 days. A prison sentence is imposed in 47 percent of the cases in the RISc-data, and the average prison term is 345 days. Of all offenders in the Prison Project, 61 percent receives a prison length at sentencing that exceeds the term of pretrial detention, with an average “extra term” of 418 days. The full prison length is approximately 370 days. Differences in the above described results are caused by differences in population and measurement between datasets. Important differences are related to the extent through which defendants have been selected through the criminal justice system: the 2007-cohort includes all suspects, the RISc-data only those with a pre-sentencing report, and the Prison Project includes pretrial detained offenders only. Consequently, when the results of ethnic disparities in sentencing are presented, it is especially valuable to pay attention to differences within datasets rather than between. The dichotomous sentencing outcomes are modeled with logistic regression. For those incarcerated, sentence length outcomes are transformed logarithmically and modeled with OLS regression. The log transformation normalizes the skewed distri- bution and addresses the fact that additional days of incarceration may have different meaning for different sentence lengths. Also, it allows for more convenient interpreta- tion of the sentence lengths in terms of their proportional increase associated with a unit increase in each explanatory variable. The regression coefficients are exponenti- ated to provide for proportional increases in the logged sentence length models.