SPSS Homework

CRJ 5083 - Spring 2019

Using the homicide sentencing dataset in blackboard, you are going to conduct a regression analysis in SPSS to determine what legal and extra-legal variables are predictors of sentence lengths (SENTTOTAL) for those convicted of homicide. Based on prior research, we believe that age (DAGE), defendant gender (SEX), whether there was victim provocation (PROVOKE), the number of victims (NUMVICT), and whether they had a trial by judge or jury (TRIALTYPE) will influence the length of the sentence handed out.

Write up the findings from the SPSS output in paragraph format. Make sure to report and interpret the r-squared value, unstandardized and standardized regression coefficients for all variables (direction, change in Y given X, significance – see below for examples), and discuss the relative effects for any significant predictors.

Instructions:

1. Run regression model in SPSS. Save and hand in your regression model output.

2. Follow the 5 basic steps for interpreting regression output. In a word document, label these 1 through 5 and report the results from the output. Also check the collinearilty diagnostics to assess any multicollinearity among the independent variables and report how you know you have met this assumption.

Important notes in reporting regression results:

Make sure to interpret the R Square value as the percent of the variance explained in the outcome by all variables included in the model and report the F-statistic when reporting significance.

Example: Age, sex, vsex, and trialtype were included in the model predicting sentence length. Together these predictors explained 33% of the variance in the sentence length handed out by the judge in the case, R2 = .33, F = 5.67, p = .03.

When interpreting the unstandardized regression coefficients make sure to provide the proper interpretation.

Example: Defendant age had a significant negative impact on sentence length (b = -1.65, p = 002). Specifically, for every one unit increase in age of the defendant, there was a 1.65 month decrease in sentence length.

Make sure to interpret the standardized regression coefficients in terms of standard deviation units and to report the standardized regression coefficients when discussing the relative magnitude of their effects on the outcome:

Example: Although both were significant predictors of delinquency, the effect of number of victims (b* = -.51, p = .002) was stronger than the effect for age (b* = .23, p = .03). That is, for every one standard deviation increase in the number of victims there was a .51 standard deviation unit decrease in sentence length compared to only a .23 standard deviation increase associated with a one unit standard deviation increase for age.

CODEBOOK for homicide dataset

SENTTotal – length of sentence received in months

dage – continuous variable measuring defendants age

sex – dichotomous variable for sex of defendant (0=female; 1=male)

provoke – dichotomous variable for victim provocation (0=no; 1=yes)

numvict – continuous variable measuring the number of homicide victims

trialtype – dichotomous variable measuring trial by judge or jury (1=judge; 2=jury)