Tuesday, September 4, 2012

Grading Research Proposals - What I'm Looking for

Hi, students. Here's a link to a video I made of a graded Senior Seminar research proposal: http://www.youtube.com/watch?v=mb10REmWNIE

This is pretty typical of the corrections I make to student papers, and I hope that by watching this video (5 minutes long) it will sensitize you to the need to write clearly, use correct words, use proper spelling and grammar, etc.

Good writing is important! It can get you into graduate school, as well as the career you want. Poor or mediocre writing will lead to graduate school/law school rejections, and competitors for the job you want getting that position instead of you. Boo! You may be very smart, but if you don't express yourself well in writing, no one will know how smart you really are.

The GOOD news is, poor writing is fairly easy to clean up. Some suggestions are: print out your paper and read it aloud on paper. Don't read off the computer screen. Also, ask a friend or fellow student to read your work, preferably out loud. Sometimes phrasing sounds clearer in your head that it does on paper. Getting another person's feedback pre-submission to me can help.

What makes good proposal writing for Seminar? Some general suggestions would be:

(1) Clearly express your ideas. Don't be vague.
(2) Follow the instructions in the syllabus.
(3) Use proper grammar and spelling. You'll never get an "A" in my course if you don't spell properly and use good grammar.

Common mistakes to avoid:

(1) Run-on sentences
(2) Sentence fragments
(3) Subject-verb agreement problems. Example: Students study for the exam, NOT students studies for the exam. (Nooo!) Also: The student studies for the exam (yes!). Not: The student study for the exam (no!).
(4) Text that is quoted without including the page number from the source material.
(5) Misuse of the apostrophe
(6) Not ending sentences with period, or other appropriate punctuation. End sentences with periods. End questions with question marks.
(7) Spelling mistakes. When in doubt, consult the dictionary.
(8) Sentences that are unclear and don't make sense
(9) Using the wrong word. There's an example of this in the video: "participates" is used when it should be "participants." Two common incorrect words I see in proposals:
-Lots of researches have found.... COMMENT: It's either research or researchers.
-I conducted a convenient sample.... COMMENT: It's a convenience sample.

Again, the link to the video: http://www.youtube.com/watch?v=mb10REmWNIE. Hope it helps, and good luck with your proposal writing.

CHW



Wednesday, July 4, 2012

Dependent Variable Survey Questions

Seminar students,

If you are planning on doing the survey methodology, consider including a 'pure' dependent variable question (or questions). Remember - you can't ask any personal questions about what your subjects have done or experienced, so you need to stick with perception and opinion questions. Here are some examples:

(1) How many youth would you say occasionally engage in delinquent acts (meaning they break the law)?
___1 in 10 ___1 in 50 ___1 in 100 ___1 in 250


(2) How many youth would you say repeatedly engage in delinquent acts?
___1 in 10 ___1 in 50 ___1 in 100 ___1 in 250

(3) How many youth would you say engage in serious, violent delinquency? 
  ___1 in 10 ___1 in 50 ___1 in 100 ___1 in 250

OR:

(4) How many sex offenders would you say reoffend (commit another sex crime) after they are released from prison? ___1 in 10 ___1 in 50 ___1 in 100 ___1 in 250

(5) How many sex offenders would you say reoffend - specifically, commit another non-sex crime such as shoplifting or using drugs - after they are released from prison?
___1 in 10 ___1 in 50 ___1 in 100 ___1 in 250

OR:

(6) How many youth would you say occasionally physically bully other kids (e.g., push, shove, punch)? ___1 in 10 ___1 in 50 ___1 in 100 ___1 in 250

(7) How many youth would you say frequently physically bully other kids?
___1 in 10 ___1 in 50 ___1 in 100 ___1 in 250

(8) How many youth would you say are occasionally physically bullied by other kids (e.g., pushed, shoved, punched)?
___1 in 10 ___1 in 50 ___1 in 100 ___1 in 250

(9) How many youth would you say are repeatedly physically bullied by other kids?
___1 in 10 ___1 in 50 ___1 in 100 ___1 in 250

This question could also be tweaked to ask about other bullying types (verbal, emotional). You could also ask a short, open-ended question that asks about what specific behaviors they think constitute "delinquency" or "bullying" (for instance). Example:

(10) How would you describe bullying? What type of behaviors constitute bullying? ______________________________________________________________________________

The type of behavior you ask about will obviously depend on what your research question is.

Thursday, April 12, 2012

Youtube Links about SPSS Analysis

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=0T-6bcS0tss&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 t-test: http://www.youtube.com/watch?v=vtkW1il7kRc&feature=BFa&list=ULrM-yPWI6zgo&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=jf-SIOFUuEo.

(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.

Results Skeleton

Your Results section will include four sections:

(1) Sample Demographics (1-3 tables; frequencies and/or descriptives)
(2) Dependent Variable(s) (1-2 tables; frequencies and/or descriptives)
(3) Independent Variables (1-3 tables; frequencies and/or descriptives)
(4) Significance test(s) (1-2 tables; correlation, independent samples t-test, 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 non-CJ 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 t-test (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, Analyze-Correlate-Bivariate. 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. (2-tailed)
.153

N
79
X2_Race/Ethnicity #
Pearson Correlation
-.020

Sig. (2-tailed)
.859

N
80
X3_Education #
Pearson Correlation
-.038

Sig. (2-tailed)
.740

N
80
X4_Enviorment #
Pearson Correlation
-.018

Sig. (2-tailed)
.874

N
80
X5_Gender #
Pearson Correlation
.014

Sig. (2-tailed)
.902

N
80
X6_Alt.Program #
Pearson Correlation
.101

Sig. (2-tailed)
.375

N
80
X7_Violence #
Pearson Correlation
.070

Sig. (2-tailed)
.539

N
80

**. Correlation is significant at the 0.01 level (2-tailed).
*. Correlation is significant at the 0.05 level (2-tailed).

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. (2-tailed)
.000
.000

N
588
588
X4_ALCOHOL INVOLVED IN CRIME
Pearson Correlation
.215(**)
.221(**)

Sig. (2-tailed)
.001
.001

N
239
239

**. Correlation is significant at the 0.01 level (2-tailed).

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 non-small (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, analyze-descriptives-crosstabs. 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%
Χ2 = .00, df= 1, p = 1.00 (take this info from the table below, and then you can delete that table)


                                                                         Chi-Square Tests


Value
df
Asymp. Sig. (2-sided)
Exact Sig. (2-sided)
Exact Sig. (1-sided)
Pearson Chi-Square
.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
Linear-by-Linear 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%
Live-in 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 t-test (appropriate for 1 continuous and 1 categorical variable). In SPSS, analyze-compare means-independent samples t-test. (See also ch. 6 in Cronk SPSS book, depending on your edition of the book). Click the categorical variable (ex: major CJ vs. non-CJ) into the grouping variable box, then “define” the grouping variable as per which number means which response: Ex: group 1 = CJ majors, group 2 = non-CJ 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

  


NYT-1 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 t-test analysis. I also ran a series of t-test 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 t-values 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 t-test 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 non-abused/non-neglected control group.