For the first three sections of your Results section (i.e., demographics, dependent variable, independent variables and covariates):
Frequencies: http://www.youtube.com/watch?v=outMas3NrQ8&feature=related (scroll ahead to ~ minute 1:25)
Descriptives: http://www.youtube.com/watch?v=0T6bcS0tss&feature=mfu_in_order&list=UL
For the fourth section of your Results section (i.e., significance tests):
The type of tests you choose to run to test for significance will depend on the type of variables you are using, continuous or categorical. Here's a video about variable types: http://www.youtube.com/watch?v=CNW9txOMYAg.
(1) Independent samples ttest: http://www.youtube.com/watch?v=vtkW1il7kRc&feature=BFa&list=ULrMyPWI6zgo&lf=mfu_in_order
(2) Correlation: http://www.youtube.com/watch?v=cNrnSEWKJgg&feature=related and also http://www.youtube.com/watch?v=Bl62NKXAe8w&feature=BFa&list=ULkMpty0exwHo&lf=mfu_in_order
Here also is a video on how to interpret the correlation coefficient (significance, sign, and size!): http://www.youtube.com/watch?v=jfSIOFUuEo.
(3) Crosstab w/chi square: http://www.youtube.com/watch?v=Ahs8jS5mJKk
Specifying missing values in SPSS: http://www.youtube.com/watch?v=wfwUC0XSnUA This is particularly relevant for the students using existing data. If the data's principal investigator(s) didn't already designate values as discrete missing values, you'll need to do this.
Thursday, April 12, 2012
Results Skeleton
Your Results section will include four sections:
(1) Sample Demographics (13 tables; frequencies and/or descriptives)
(2) Dependent Variable(s) (12 tables; frequencies and/or descriptives)
(3) Independent Variables (13 tables; frequencies and/or descriptives)
(4) Significance test(s) (12 tables; correlation, independent samples ttest, and/or crosstab w/chi square tests).
Note: the tables shown below were copy/pasted from Word, and do not come out well in this blog (i.e., they're all smooshed). Refer to the Word document I emailed to everyone for examples of presentable SPSS output tables.
SECTION 1: Sample Demographics
Table 1: Newspapers from which sample comes
As seen in Table 1, the sample was split nearly evenly between articles from the NY Times (51%) and the LA Times (49%).
Table 2: Sample Gender
As seen above in Table 2, my sample heavily consisted of slightly more female (58%) than male (42%) students. My sample was split about evenly between CJ majors (52%) and nonCJ majors (48%; not shown in table format).
SECTION 2: Dependent Variables
Table 3: Dependent Variable
As seen above in Table 3, my dependent variable – number of mentions of recidivism – was mentioned on average .44 times in the newspaper articles. There was a minimum of 0 mentions in one or more articles, and a maximum mention of 5 in 1 or more articles. AND DON’T FORGET TO ADD IN YOUR QUALITATIVE DATA (THE WORDS AND PHRASES). Examples of the words and phrases that mean recidivism are: chronic offender, repeat offenders, did it again, and went on to commit more crime.
Other Table option (for survey question dependent variables)
As seen above in Other Table 3, nearly 84 percent of my sample thought that relapse by drug users was fairly common, with either 1 in 10 or 1 in 25 individuals using drugs again. Fifteen percent of my sample felt that only 1 in 50 drug users would relapse, and 1 individual felt that drug relapse was rare (i.e., only 1 in 100 users).
SECTION 3: Independent Variables
Table 4: Independent Variables, Causal Factors of Recidivism
As seen above in Table 4, of the independent variables, number of mentions of age and violence appeared the most frequently. Age was mentioned an average of 9 times in one or more articles, with a minimum of 1 mention and a maximum of 35 mentions. Violence was mentioned on average 6 times in the articles. Violence had a minimum of 0 mentions, and a maximum of 30 mentions in 1 or more articles. AND DON’T FORGET TO ADD IN YOUR QUALITATIVE DATA (THE WORDS AND PHRASES) Language used to describe violence included: rape, assault, beatings, violence, abuse, homicide, robbing, and drowning.
SECTION 4: Significance testing for testing your research question or hypothesis)
• Option 1: Correlation (appropriate for 2 continuous variables), OR
• Option 2: crosstab w/chi square (appropriate for 2 categorical variables) OR
• Option 3: independent samples ttest (appropriate for 1 continuous and 1 categorical variable).
You’ll run 1 or 2 of these. Which tests you run will depend on the type of variables you are using, in particular continuous or categorical variables. See this video (skip the add) re: variable types: http://www.youtube.com/watch?v=CNW9txOMYAg.
Recall: Categorical variables have categories of responses, like male/female, yes/no, CNN.com/FoxNews.com, strongly agree…strongly disagree
Continuous variables do NOT have categories of responses; ex: age, number of mentions of whatever (delinquency, joining gangs, family structure, IPV, etc.)
Option 1: Correlation (appropriate for 2 continuous variables); in SPSS, AnalyzeCorrelateBivariate. Click over your continuous variables into the “variables” box, beginning with the dependent variable on top. See also ch. 5 (I think, depending on your edition of the book), “Pearson Correlation Coefficient” in the Cronk SPSS book. Also see this video clip:
**. Correlation is significant at the 0.01 level (2tailed).
*. Correlation is significant at the 0.05 level (2tailed).
As seen above in Table 5, a correlation analysis showed that none of my independent variables were significantly related to my dependent variable. I was surprised to learn that recidivism and environment are not related to each other. This is counter to what I hypothesized.
Table 7: Relationship of Number of Offenses, Age, and Alcohol Use
**. Correlation is significant at the 0.01 level (2tailed).
As seen above in Table 7, there were statistically significant relationships between both of my dependent variables  number of juvenile and adult offenses  and both independent variables. Age at first offense was negatively related to both dependent variables, suggesting that the younger the age of beginning criminal involvement, the greater the number of both juvenile and adult offenses. Note that the size of the correlation coefficient was particularly robust (big) for X3 and Y1 (juvenile offenses). Alcohol use during the crime was positively related to both dependent variables, suggesting that the more an offender uses alcohol, the more likely they are to commit a crime. There was also a significant, positive, and nonsmall (r = .3) relationship between Y1 and Y2, not shown in table format.
Option 2: crosstab w/chi square (appropriate for 2 categorical variables). In SPSS, analyzedescriptivescrosstabs. Click dependent variable into either row or column box, and independent variable into the other box (whatever you didn’t click the dependent variable into). In the “statistics” tab click “chi square” test. In “cells” tab, select row percents. See also Cronk book this is ch. 7 (I think, depending on your edition of the book). See also this video clip.
Table 8: Subject Gender by Treatment or Control Group (Cathy Widom data)
As seen above in Table 8, there was no pattern of difference in which gender was assigned to which experimental group. In fact, identical numbers of females AND males appear in both the treatment (subject) and control groups (e.g., 462 females in both the subject and control groups). Thus, the chi square value is not statistically significant.
Table 9: Marital Status of Male and Female Police Officers
As seen above in Table 9, different marital and cohabitation patterns were evident for male vs. female police officers. Male officers were more likely to be married (64% of the sample of 938 officers), compared with only 36% of the female officers (out the n=154). Female officers were twice as likely to cohabitate with an intimate partner than were male officers (14% vs. 7%, respectively). More female officers were divorced (22%) than were male officers (just under 11%). More female officers were also single (27%) than their male counterparts (18%). The results of the chi square test were statistically significant. These results suggest that policing as a profession may be more detrimental to female officers' personal/romantic life and stability than it (policing) is for male officers. This may be due to... (here you'd speculate, as appropriate to your research question, on why fewer female officers get or remain married... social expectations of wives, etc.)
Option 3: independent samples ttest (appropriate for 1 continuous and 1 categorical variable). In SPSS, analyzecompare meansindependent samples ttest. (See also ch. 6 in Cronk SPSS book, depending on your edition of the book). Click the categorical variable (ex: major CJ vs. nonCJ) into the grouping variable box, then “define” the grouping variable as per which number means which response: Ex: group 1 = CJ majors, group 2 = nonCJ majors. In the “test variable” box click over the dependent variable. Click okay to run. See also this video clip.
Table 11: Comparison of Mean Mentions of Recidivism by News Source
As seen above in Table 11, recidivism and its synonyms were mentioned more frequently in the NYT than in the LA Times. In the NY Times, recidivism was mentioned on average .68 times in the articles, compared with the very low .18 mean mentions in the LA Times. This difference was statistically significant (p=.02) as per the ttest analysis. I also ran a series of ttest analyses of all my independent variables by which news source (NYT or LAT), but none of these were statistically significant.
Table 12: Comparison of Mean Juvenile and Adult Offenses by Subject or Control Group
As seen above in Table 12, individuals in the "subject" group committed higher mean (average) numbers of both juvenile and adult offenses than did the "control" individuals. The ttest results for both dependent variables were statistically significant. The differences in means for the adult offenses (Y2) were particularly noticeable: a mean of over 9 offenses for the subjects, compared with less than 5 offenses on average for the controls. These findings suggest that the "subject" group  those individuals subjected to severe physical or sexual abuse or neglect in childhood  go on to break the law more than do the nonabused/nonneglected control group.
(1) Sample Demographics (13 tables; frequencies and/or descriptives)
(2) Dependent Variable(s) (12 tables; frequencies and/or descriptives)
(3) Independent Variables (13 tables; frequencies and/or descriptives)
(4) Significance test(s) (12 tables; correlation, independent samples ttest, and/or crosstab w/chi square tests).
Note: the tables shown below were copy/pasted from Word, and do not come out well in this blog (i.e., they're all smooshed). Refer to the Word document I emailed to everyone for examples of presentable SPSS output tables.
SECTION 1: Sample Demographics
Table 1: Newspapers from which sample comes

Frequency

Percent

Valid Percent

Cumulative Percent


Valid

New York
Times (1)

41

51.3

51.3

51.3

Los
Angeles Times (2) 
39

48.8

48.8

100.0


Total 
80

100.0

100.0


As seen in Table 1, the sample was split nearly evenly between articles from the NY Times (51%) and the LA Times (49%).
Table 2: Sample Gender
Frequency

Percent

Valid Percent

Cumulative Percent


Valid

male (1)

34

42.5

42.5

42.5

female
(2) 
46

57.5

57.5

100.0


Total 
80

100.0

100.0

As seen above in Table 2, my sample heavily consisted of slightly more female (58%) than male (42%) students. My sample was split about evenly between CJ majors (52%) and nonCJ majors (48%; not shown in table format).
SECTION 2: Dependent Variables
Table 3: Dependent Variable
N

Minimum

Maximum

Mean

Std. Deviation


Y_Recidivism
#: number of mentions of recidivism (dep. var.)

80

0

5

.44

.992

Valid N
(listwise)

80

As seen above in Table 3, my dependent variable – number of mentions of recidivism – was mentioned on average .44 times in the newspaper articles. There was a minimum of 0 mentions in one or more articles, and a maximum mention of 5 in 1 or more articles. AND DON’T FORGET TO ADD IN YOUR QUALITATIVE DATA (THE WORDS AND PHRASES). Examples of the words and phrases that mean recidivism are: chronic offender, repeat offenders, did it again, and went on to commit more crime.
Other Table option (for survey question dependent variables)
Table
3:
How many individuals with drug use problem who
attempt to stop end up relapsing?
Frequency

Percent

Valid Percent

Cumulative Percent


Valid

1 in 10
(1)

36

45.0

45.0

45.0

1 in 25
(2) 
31

38.8

38.8

83.8


1 in 50
(3) 
12

15.0

15.0

98.8


1 in 100
(4) 
1

1.3

1.3

100.0


Total 
80

100.0

100.0

As seen above in Other Table 3, nearly 84 percent of my sample thought that relapse by drug users was fairly common, with either 1 in 10 or 1 in 25 individuals using drugs again. Fifteen percent of my sample felt that only 1 in 50 drug users would relapse, and 1 individual felt that drug relapse was rare (i.e., only 1 in 100 users).
SECTION 3: Independent Variables
Table 4: Independent Variables, Causal Factors of Recidivism
N

Minimum

Maximum

Mean

Std. Deviation


X1_Age #

79

1

35

9.27

7.074

X2_Race/Ethnicity
#

80

0

22

1.03

3.085

X3_Education
#

80

0

7

1.46

2.098

X4_Enviorment
#

80

0

6

1.33

1.682

X5_Gender
#

80

0

17

1.96

3.021

X6_Alt.Program
#

80

0

14

1.99

2.848

X7_Violence
#

80

0

30

6.39

5.568

Valid N
(listwise)

79

As seen above in Table 4, of the independent variables, number of mentions of age and violence appeared the most frequently. Age was mentioned an average of 9 times in one or more articles, with a minimum of 1 mention and a maximum of 35 mentions. Violence was mentioned on average 6 times in the articles. Violence had a minimum of 0 mentions, and a maximum of 30 mentions in 1 or more articles. AND DON’T FORGET TO ADD IN YOUR QUALITATIVE DATA (THE WORDS AND PHRASES) Language used to describe violence included: rape, assault, beatings, violence, abuse, homicide, robbing, and drowning.
SECTION 4: Significance testing for testing your research question or hypothesis)
• Option 1: Correlation (appropriate for 2 continuous variables), OR
• Option 2: crosstab w/chi square (appropriate for 2 categorical variables) OR
• Option 3: independent samples ttest (appropriate for 1 continuous and 1 categorical variable).
You’ll run 1 or 2 of these. Which tests you run will depend on the type of variables you are using, in particular continuous or categorical variables. See this video (skip the add) re: variable types: http://www.youtube.com/watch?v=CNW9txOMYAg.
Recall: Categorical variables have categories of responses, like male/female, yes/no, CNN.com/FoxNews.com, strongly agree…strongly disagree
Continuous variables do NOT have categories of responses; ex: age, number of mentions of whatever (delinquency, joining gangs, family structure, IPV, etc.)
Option 1: Correlation (appropriate for 2 continuous variables); in SPSS, AnalyzeCorrelateBivariate. Click over your continuous variables into the “variables” box, beginning with the dependent variable on top. See also ch. 5 (I think, depending on your edition of the book), “Pearson Correlation Coefficient” in the Cronk SPSS book. Also see this video clip:
Y_Recidivism #: number of mentions of recidivism (dep. var.)


X1_Age #

Pearson
Correlation

.162

Sig.
(2tailed)

.153


N

79


X2_Race/Ethnicity
#

Pearson
Correlation

.020

Sig.
(2tailed)

.859


N

80


X3_Education
#

Pearson
Correlation

.038

Sig.
(2tailed)

.740


N

80


X4_Enviorment
#

Pearson
Correlation

.018

Sig.
(2tailed)

.874


N

80


X5_Gender
#

Pearson
Correlation

.014

Sig.
(2tailed)

.902


N

80


X6_Alt.Program
#

Pearson
Correlation

.101

Sig.
(2tailed)

.375


N

80


X7_Violence
#

Pearson
Correlation

.070

Sig.
(2tailed)

.539


N

80

**. Correlation is significant at the 0.01 level (2tailed).
*. Correlation is significant at the 0.05 level (2tailed).
As seen above in Table 5, a correlation analysis showed that none of my independent variables were significantly related to my dependent variable. I was surprised to learn that recidivism and environment are not related to each other. This is counter to what I hypothesized.
Table 7: Relationship of Number of Offenses, Age, and Alcohol Use
Y1_NUMBER OF JUVENILE OFFENSES

Y2_NUMBER OF ADULT OFFENSES


X3_RECODE
OF AGE AT 1ST OFFENSE

Pearson
Correlation

.626(**)

.335(**)

Sig.
(2tailed)

.000

.000


N

588

588


X4_ALCOHOL
INVOLVED IN CRIME

Pearson
Correlation

.215(**)

.221(**)

Sig.
(2tailed)

.001

.001


N

239

239

**. Correlation is significant at the 0.01 level (2tailed).
As seen above in Table 7, there were statistically significant relationships between both of my dependent variables  number of juvenile and adult offenses  and both independent variables. Age at first offense was negatively related to both dependent variables, suggesting that the younger the age of beginning criminal involvement, the greater the number of both juvenile and adult offenses. Note that the size of the correlation coefficient was particularly robust (big) for X3 and Y1 (juvenile offenses). Alcohol use during the crime was positively related to both dependent variables, suggesting that the more an offender uses alcohol, the more likely they are to commit a crime. There was also a significant, positive, and nonsmall (r = .3) relationship between Y1 and Y2, not shown in table format.
Option 2: crosstab w/chi square (appropriate for 2 categorical variables). In SPSS, analyzedescriptivescrosstabs. Click dependent variable into either row or column box, and independent variable into the other box (whatever you didn’t click the dependent variable into). In the “statistics” tab click “chi square” test. In “cells” tab, select row percents. See also Cronk book this is ch. 7 (I think, depending on your edition of the book). See also this video clip.
Table 8: Subject Gender by Treatment or Control Group (Cathy Widom data)
X1_SUBJECT / CONTROL

Total


Subjects 
Controls


X2_CHILD'S
GENDER

Female

Count

462

462

924

% within
X1_SUBJECT / CONTROL 
52.7%

52.7%

52.7%


Male 
Count

415

415

830


% within
X1_SUBJECT / CONTROL 
47.3%

47.3%

47.3%


Total

Count

877

877

1754


% within
X1_SUBJECT / CONTROL 
100.0%

100.0%

100.0%

Î§2 = .00, df= 1, p = 1.00
(take this info from the table below, and then you can delete that table)
ChiSquare
Tests
Value

df

Asymp. Sig. (2sided)

Exact Sig. (2sided)

Exact Sig. (1sided)


Pearson ChiSquare

.000(b)

1

1.000


Continuity
Correction(a)

.000

1

1.000


Likelihood
Ratio

.000

1

1.000


Fisher's
Exact Test

1.000

.519


LinearbyLinear
Association

.000

1

1.000


N of
Valid Cases

1754

a Computed only for a 2x2 table
b 0 cells (.0%) have expected count less than
5. The minimum expected count is 415.00.
As seen above in Table 8, there was no pattern of difference in which gender was assigned to which experimental group. In fact, identical numbers of females AND males appear in both the treatment (subject) and control groups (e.g., 462 females in both the subject and control groups). Thus, the chi square value is not statistically significant.
Table 9: Marital Status of Male and Female Police Officers
GENDER

Total


Male 
Female


Y1_MARITAL
STATUS

Married

Count

601

56

657

% within
GENDER 
64.1%

36.4%

60.2%


Livein partner 
Count

66

22

88


% within
GENDER 
7.0%

14.3%

8.1%


Divorced/Separated 
Count

100

34

134


% within
GENDER 
10.7%

22.1%

12.3%


Single 
Count

171

42

213


% within
GENDER 
18.2%

27.3%

19.5%


Total

Count

938

154

1092


% within
GENDER 
100.0%

100.0%

100.0%

Chi square: 45.104, df=3,
p= .00
As seen above in Table 9, different marital and cohabitation patterns were evident for male vs. female police officers. Male officers were more likely to be married (64% of the sample of 938 officers), compared with only 36% of the female officers (out the n=154). Female officers were twice as likely to cohabitate with an intimate partner than were male officers (14% vs. 7%, respectively). More female officers were divorced (22%) than were male officers (just under 11%). More female officers were also single (27%) than their male counterparts (18%). The results of the chi square test were statistically significant. These results suggest that policing as a profession may be more detrimental to female officers' personal/romantic life and stability than it (policing) is for male officers. This may be due to... (here you'd speculate, as appropriate to your research question, on why fewer female officers get or remain married... social expectations of wives, etc.)
Option 3: independent samples ttest (appropriate for 1 continuous and 1 categorical variable). In SPSS, analyzecompare meansindependent samples ttest. (See also ch. 6 in Cronk SPSS book, depending on your edition of the book). Click the categorical variable (ex: major CJ vs. nonCJ) into the grouping variable box, then “define” the grouping variable as per which number means which response: Ex: group 1 = CJ majors, group 2 = nonCJ majors. In the “test variable” box click over the dependent variable. Click okay to run. See also this video clip.
Table 11: Comparison of Mean Mentions of Recidivism by News Source
NYT1
LAT 2

N

Mean

Std. Deviation

Std. Error Mean


Y_Recidivism
#: number of mentions of recidivism (dep. var.)

New York
Times (1)

41

.68

1.234

.193

Los Angeles Times (2) 
39

.18

.556

.089

T= 2.33, p= .022
As seen above in Table 11, recidivism and its synonyms were mentioned more frequently in the NYT than in the LA Times. In the NY Times, recidivism was mentioned on average .68 times in the articles, compared with the very low .18 mean mentions in the LA Times. This difference was statistically significant (p=.02) as per the ttest analysis. I also ran a series of ttest analyses of all my independent variables by which news source (NYT or LAT), but none of these were statistically significant.
Table 12: Comparison of Mean Juvenile and Adult Offenses by Subject or Control Group
X1_SUBJECT
/ CONTROL

N

Mean

Std. Deviation

Std. Error Mean


Y1_NUMBER
OF JUVENILE OFFENSES

Subjects

397

1.63

3.138

.158

Controls 
192

.35

1.097

.079


Y2_NUMBER
OF ADULT OFFENSES

Subjects

397

9.24

13.821

.694

Controls 
192

4.43

7.002

.505

Both tvalues are
statistically significant. Y1: t= 5.48, p= .00; Y2: t=4.55, p= .00
.
As seen above in Table 12, individuals in the "subject" group committed higher mean (average) numbers of both juvenile and adult offenses than did the "control" individuals. The ttest results for both dependent variables were statistically significant. The differences in means for the adult offenses (Y2) were particularly noticeable: a mean of over 9 offenses for the subjects, compared with less than 5 offenses on average for the controls. These findings suggest that the "subject" group  those individuals subjected to severe physical or sexual abuse or neglect in childhood  go on to break the law more than do the nonabused/nonneglected control group.
Subscribe to:
Posts (Atom)