everild-blog

[everild-blog]

Week 4: Testing a Potential Moderator

This week I was testing whether the amount of time adolescents spend playing video games is related to the presence of sexually related events among the first five events the participants list in their description of an ideal romantic relationship.

During one of the previous weeks (http://everild-blog.tumblr.com/post/153639207924/week-2-running-a-chi-square-test-of-independence), I found that there is, in fact, a correlation: the teenagers who played computer or video games had less sexualized perception of romantic relationships than those who did not. 

For current task, I chose biological sex of the participants as a moderator. Other variables were the same; I ran a Chi-Square test with post-hocs. 

Here is my code:

LIBNAME mydata "/courses/d1406ae5ba27fe300 " access=readonly; DATA new; set mydata.addhealth_pds;

LABEL H1DA10="Hours of playing video or computer games per week" H1ID2_1="First romantic event" H1ID2_2="Second romantic event" H1ID2_3="Third romantic event" H1ID2_4="Fourth romantic event" H1ID2_5="Fifth romantic event" BIO_SEX="Biological sex" SEX_SUM="The proportion of sexually related romantic events within the first five suggested" SEX_FIRST="The number of the first sexually related romantic event suggested" TIME_GROUP="How many minutes per day do you play video or computer games?"

IF H1IR6=0;/*indicates the absence of a third person during the interview*/

/*MAKING GROUPS BASED ON THE NUMBER OF HOURS PER WEEK OF PLAYING COMPUTER GAMES*/ IF H1DA10=996 THEN H1DA10=.; IF H1DA10=998 THEN H1DA10=.;

MINSPERDAY=H1DA10*60/7; IF MINSPERDAY=0 THEN TIME_GROUP=0; /*not playing computer games*/ ELSE IF MINSPERDAY>0 AND MINSPERDAY<=15 THEN TIME_GROUP=8; /*playing less than 15 minutes per day*/ ELSE IF MINSPERDAY> 15 AND MINSPERDAY<= 20 THEN TIME_GROUP=18; /*playing 15 to 20 minutes per day*/ ELSE IF MINSPERDAY> 20 AND MINSPERDAY<= 45 THEN TIME_GROUP=33; /*playing 20 to 45 minutes per day*/ ELSE IF MINSPERDAY>45 THEN TIME_GROUP=86; /*playing more than 45 minutes per day*/

/*DELETING THE MISSING DATA ON THE SUGGESTED ROMANTIC EVENTS*/ IF H1ID2_1>90 AND H1ID2_1 NE 97 THEN H1ID2_1=.; IF H1ID2_1=97 THEN H1ID2_1=0; /*97 encodes legitimate skip (no cards left)*/

IF H1ID2_2>90 AND H1ID2_2 NE 97 THEN H1ID2_2=.; IF H1ID2_2=97 THEN H1ID2_2=0;/*97 encodes legitimate skip (no cards left)*/

IF H1ID2_3>90 AND H1ID2_3 NE 97 THEN H1ID2_3=.; IF H1ID2_3=97 THEN H1ID2_3=0;/*97 encodes legitimate skip (no cards left)*/

IF H1ID2_4>90 AND H1ID2_4 NE 97 THEN H1ID2_4=.; IF H1ID2_4=97 THEN H1ID2_4=0;/*97 encodes legitimate skip (no cards left)*/

IF H1ID2_5>90 AND H1ID2_5 NE 97 THEN H1ID2_5=.; IF H1ID2_5=97 THEN H1ID2_5=0;/*97 encodes legitimate skip (no cards left)*/

/*DEFINING WHETHER THE EVENT WAS SEXUALLY RELATED OR NOT*/ If H1ID2_1>11 AND H1ID2_1<17 THEN SEX_1=1; ELSE IF H1ID2_1=. THEN SEX_1=.; ELSE SEX_1=0; If H1ID2_2>11 AND H1ID2_2<17 THEN SEX_2=1; ELSE IF H1ID2_2=. THEN SEX_2=.; ELSE SEX_2=0; If H1ID2_3>11 AND H1ID2_3<17 THEN SEX_3=1; ELSE IF H1ID2_3=. THEN SEX_3=.; ELSE SEX_3=0; If H1ID2_4>11 AND H1ID2_4<17 THEN SEX_4=1; ELSE IF H1ID2_4=. THEN SEX_4=.; ELSE SEX_4=0; If H1ID2_5>11 AND H1ID2_5<17 THEN SEX_5=1; ELSE IF H1ID2_5=. THEN SEX_5=.; ELSE SEX_5=0;

SEX_SUM=SUM (of SEX_1 SEX_2 SEX_3 SEX_4 SEX_5)/5; /*defines the proportion of sexual events within the first five*/ IF SEX_1=. OR SEX_2=. OR SEX_3=. OR SEX_4=. OR SEX_5=. THEN SEX_SUM=.; /*deleting entries with missing data*/ IF SEX_1=. OR SEX_2=. OR SEX_3=. OR SEX_4=. OR SEX_5=. THEN SEX_FIRST=.;/*deleting entries with missing data*/ IF SEX_SUM=0 THEN SEX=0; IF SEX_SUM>0 THEN SEX=1;

IF SEX_1=. OR SEX_2=. OR SEX_3=. OR SEX_4=. OR SEX_5=. THEN SEX_FIRST=.;/*deleting entries with missing data*/

PROC SORT; by BIO_SEX; PROC FREQ; TABLES SEX*TIME_GROUP/CHISQ; BY BIO_SEX; PROC GCHART; VBAR TIME_GROUP/discrete type=mean SUMVAR=SEX;

RUN;

DATA COMPARISON1; SET NEW; IF TIME_GROUP=8 OR TIME_GROUP=18; PROC SORT; BY AID; PROC FREQ; TABLES SEX*TIME_GROUP/CHISQ; RUN;

DATA COMPARISON2; SET NEW; IF TIME_GROUP=8 OR TIME_GROUP=33; PROC SORT; BY AID; PROC FREQ; TABLES SEX*TIME_GROUP/CHISQ; RUN;

DATA COMPARISON3; SET NEW; IF TIME_GROUP=8 OR TIME_GROUP=86; PROC SORT; BY AID; PROC FREQ; TABLES SEX*TIME_GROUP/CHISQ; RUN;

DATA COMPARISON4; SET NEW; IF TIME_GROUP=18 OR TIME_GROUP=33; PROC SORT; BY AID; PROC FREQ; TABLES SEX*TIME_GROUP/CHISQ; RUN;

DATA COMPARISON5; SET NEW; IF TIME_GROUP=18 OR TIME_GROUP=86; PROC SORT; BY AID; PROC FREQ; TABLES SEX*TIME_GROUP/CHISQ; RUN;

DATA COMPARISON6; SET NEW; IF TIME_GROUP=33 OR TIME_GROUP=86; PROC SORT; BY AID; PROC FREQ; TABLES SEX*TIME_GROUP/CHISQ; RUN;

DATA COMPARISON7; SET NEW; IF TIME_GROUP=0 OR TIME_GROUP=8; PROC SORT; BY AID; PROC FREQ; TABLES SEX*TIME_GROUP/CHISQ; RUN;

DATA COMPARISON8; SET NEW; IF TIME_GROUP=0 OR TIME_GROUP=18; PROC SORT; BY AID; PROC FREQ; TABLES SEX*TIME_GROUP/CHISQ; RUN;

DATA COMPARISON9; SET NEW; IF TIME_GROUP=0 OR TIME_GROUP=33; PROC SORT; BY AID; PROC FREQ; TABLES SEX*TIME_GROUP/CHISQ; RUN;

DATA COMPARISON10; SET NEW; IF TIME_GROUP=0 OR TIME_GROUP=86; PROC SORT; BY AID; PROC FREQ; TABLES SEX*TIME_GROUP/CHISQ; RUN;

Here is the output of the program:

It turns out to be that in both groups of teenagers (male, BIO_SEX=1, and female, BIO_SEX=2) there is a significant correlation between the studied variables. Both girls and boys who do not play video games tend to have a more sexualized perception of a romantic relationship than the “gamers”, and there is no difference between male and female adolescents. Therefore, biological sex is not a moderating variable in this case.

The post-hoc comparisons revealed significant differences (p<0.005, as I have 10 comparisons) between those who did not play video games (0 minutes per day) and all of the groups of “gamers” (playing 8, 18, 33, 86 minutes per day). There was no significant difference between various groups of “gamers”.

[everild-blog]

Week 3: Generating a Correlation Coefficient

In order to practice in calculating the correlation coefficient, I chose 2 variables from the AddHealth data set: the number of hours per week the respondent spends playing computer or video games (H1DA10, quantitative) and the year of his or her birth (H1GI1Y, designating the last two digits of the year; qualitative, ranging from 74 to 96 with a mean of 79).

Here is my code:

LIBNAME mydata "/courses/d1406ae5ba27fe300 " access=readonly; DATA new; set mydata.addhealth_pds;

LABEL H1DA10="Hours of playing video or computer games per week" H1GI1Y="Year of birth";

IF H1IR6=0;/*indicates the absence of a third person during the interview*/

PROC SORT; by AID;/*sorting by the unique identifier*/ PROC CORR; VAR H1GI1Y H1DA10;

RUN;

I got the following results:

Although the p-value is small enough in this case (p<0.0001), the value of the correlation coefficient is pretty close to zero (r=0.10255), indicating a weak, yet statistically significant correlation. As r>0, we can infer that this correlation is positive, i.e., the higher is the value of the year of birth (ergo, the younger is the respondent), the more time he or she spends playing computer or video games per week. 

[everild-blog]

Week 2: running a Chi-Square Test of Independence

cThis week I was testing whether the amount of time adolescents spend playing video games is related to the presence of sexually related events among the first five events the participants list in their description of an ideal romantic relationship.

My explanatory variable was TIME_GROUP (how many minutes per day a participant plays video or computer games). It is categorical, with 5 possible values: 0; 8; 18; 33; 86 (minutes/day).

My response variable was SEX (indicating presence or absence of sexually related events among the first five listed in the description of an ideal romantic relationship; you can get more information in these posts: initial definition of the variables, creation of categorical variables). SEX was set as 0, if SEX_SUM was equal to 0 (i.e., there were no sexually related events among the first five), and SEX was set to 1, if SEX_SUM was not equal to 0 (i.e., there was at least 1 sexually related event). I had to introduce a new categorical variable, as my response variable was to have only 1 df (i.e., have only 2 possible values) for the Chi-Square test.

As usually, I included only the cases when no third person was present during the interview (IF H1IR6=0;).

Here is my code:

LIBNAME mydata "/courses/d1406ae5ba27fe300 " access=readonly; DATA new; set mydata.addhealth_pds;

LABEL H1DA10="Hours of playing video or computer games per week" H1ID2_1="First romantic event" H1ID2_2="Second romantic event" H1ID2_3="Third romantic event" H1ID2_4="Fourth romantic event" H1ID2_5="Fifth romantic event" BIO_SEX="Biological sex" SEX_SUM="The proportion of sexually related romantic events within the first five suggested" SEX_FIRST="The number of the first sexually related romantic event suggested" TIME_GROUP="How many minutes per day do you play video or computer games?";

IF H1IR6=0;/*indicates the absence of a third person during the interview*/

/*MAKING GROUPS BASED ON THE NUMBER OF HOURS PER WEEK OF PLAYING COMPUTER GAMES*/ IF H1DA10=996 THEN H1DA10=.; IF H1DA10=998 THEN H1DA10=.;

MINSPERDAY=H1DA10*60/7; IF MINSPERDAY=0 THEN TIME_GROUP=0; /*not playing computer games*/ ELSE IF MINSPERDAY>0 AND MINSPERDAY<=15 THEN TIME_GROUP=8; /*playing less than 15 minutes per day*/ ELSE IF MINSPERDAY> 15 AND MINSPERDAY<= 20 THEN TIME_GROUP=18; /*playing 15 to 20 minutes per day*/ ELSE IF MINSPERDAY> 20 AND MINSPERDAY<= 45 THEN TIME_GROUP=33; /*playing 20 to 45 minutes per day*/ ELSE IF MINSPERDAY>45 THEN TIME_GROUP=86; /*playing more than 45 minutes per day*/

/*DELETING THE MISSING DATA ON THE SUGGESTED ROMANTIC EVENTS*/ IF H1ID2_1>90 AND H1ID2_1 NE 97 THEN H1ID2_1=.; IF H1ID2_1=97 THEN H1ID2_1=0; /*97 encodes legitimate skip (no cards left)*/

IF H1ID2_2>90 AND H1ID2_2 NE 97 THEN H1ID2_2=.; IF H1ID2_2=97 THEN H1ID2_2=0;/*97 encodes legitimate skip (no cards left)*/

IF H1ID2_3>90 AND H1ID2_3 NE 97 THEN H1ID2_3=.; IF H1ID2_3=97 THEN H1ID2_3=0;/*97 encodes legitimate skip (no cards left)*/

IF H1ID2_4>90 AND H1ID2_4 NE 97 THEN H1ID2_4=.; IF H1ID2_4=97 THEN H1ID2_4=0;/*97 encodes legitimate skip (no cards left)*/

IF H1ID2_5>90 AND H1ID2_5 NE 97 THEN H1ID2_5=.; IF H1ID2_5=97 THEN H1ID2_5=0;/*97 encodes legitimate skip (no cards left)*/

/*DEFINING WHETHER THE EVENT WAS SEXUALLY RELATED OR NOT*/ If H1ID2_1>11 AND H1ID2_1<17 THEN SEX_1=1; ELSE IF H1ID2_1=. THEN SEX_1=.; ELSE SEX_1=0; If H1ID2_2>11 AND H1ID2_2<17 THEN SEX_2=1; ELSE IF H1ID2_2=. THEN SEX_2=.; ELSE SEX_2=0; If H1ID2_3>11 AND H1ID2_3<17 THEN SEX_3=1; ELSE IF H1ID2_3=. THEN SEX_3=.; ELSE SEX_3=0; If H1ID2_4>11 AND H1ID2_4<17 THEN SEX_4=1; ELSE IF H1ID2_4=. THEN SEX_4=.; ELSE SEX_4=0; If H1ID2_5>11 AND H1ID2_5<17 THEN SEX_5=1; ELSE IF H1ID2_5=. THEN SEX_5=.; ELSE SEX_5=0;

SEX_SUM=SUM (of SEX_1 SEX_2 SEX_3 SEX_4 SEX_5)/5; /*defines the proportion of sexual events within the first five*/ IF SEX_1=. OR SEX_2=. OR SEX_3=. OR SEX_4=. OR SEX_5=. THEN SEX_SUM=.; /*deleting entries with missing data*/ IF SEX_1=. OR SEX_2=. OR SEX_3=. OR SEX_4=. OR SEX_5=. THEN SEX_FIRST=.;/*deleting entries with missing data*/ IF SEX_SUM=0 THEN SEX=0; IF SEX_SUM>0 THEN SEX=1;

PROC SORT; by AID;/*sorting by the unique identifier*/ PROC FREQ; TABLES SEX*TIME_GROUP/CHISQ;

RUN;

DATA COMPARISON1; SET NEW; IF TIME_GROUP=8 OR TIME_GROUP=18; PROC SORT; BY AID; PROC FREQ; TABLES SEX*TIME_GROUP/CHISQ; RUN;

DATA COMPARISON2; SET NEW; IF TIME_GROUP=8 OR TIME_GROUP=33; PROC SORT; BY AID; PROC FREQ; TABLES SEX*TIME_GROUP/CHISQ; RUN;

DATA COMPARISON3; SET NEW; IF TIME_GROUP=8 OR TIME_GROUP=86; PROC SORT; BY AID; PROC FREQ; TABLES SEX*TIME_GROUP/CHISQ; RUN;

DATA COMPARISON4; SET NEW; IF TIME_GROUP=18 OR TIME_GROUP=33; PROC SORT; BY AID; PROC FREQ; TABLES SEX*TIME_GROUP/CHISQ; RUN;

DATA COMPARISON5; SET NEW; IF TIME_GROUP=18 OR TIME_GROUP=86; PROC SORT; BY AID; PROC FREQ; TABLES SEX*TIME_GROUP/CHISQ; RUN;

DATA COMPARISON6; SET NEW; IF TIME_GROUP=33 OR TIME_GROUP=86; PROC SORT; BY AID; PROC FREQ; TABLES SEX*TIME_GROUP/CHISQ; RUN;

DATA COMPARISON7; SET NEW; IF TIME_GROUP=0 OR TIME_GROUP=8; PROC SORT; BY AID; PROC FREQ; TABLES SEX*TIME_GROUP/CHISQ; RUN;

DATA COMPARISON8; SET NEW; IF TIME_GROUP=0 OR TIME_GROUP=18; PROC SORT; BY AID; PROC FREQ; TABLES SEX*TIME_GROUP/CHISQ; RUN;

DATA COMPARISON9; SET NEW; IF TIME_GROUP=0 OR TIME_GROUP=33; PROC SORT; BY AID; PROC FREQ; TABLES SEX*TIME_GROUP/CHISQ; RUN;

DATA COMPARISON10; SET NEW; IF TIME_GROUP=0 OR TIME_GROUP=86; PROC SORT; BY AID; PROC FREQ; TABLES SEX*TIME_GROUP/CHISQ; RUN;

The Chi-Square test revealed that the time spent playing computer or video games and the presence of sexually related events in the description of a perfect romantic relationship were significantly associated (X2=33.6297; df=4; p< .0001).

The post-hoc comparisons revealed significant differences (p<0.005, as I have 10 comparisons) between those who did not play video games (0 minutes per day) and all of the groups of “gamers” (playing 8, 18, 33, 86 minutes per day). The teenagers who played computer or video games had less sexualized perception of romantic relationships than those who did not. There was no significant difference between various groups of “gamers”.

I was also interested in conducting the same tests on the sex-based subsets of teenagers (male or female only). In order to do that, I used another variable, BIO_SEX, and repeated the analysis.

It turned out to be that, although the Chi-Square test showed significant differences in case of male teenagers (X2=10.7119; df=4; p=0.0300), the post-hoc tests did not confirm any differences. Among the female teenagers (X2=18.4546; df=4; p=0.0010), the post-hoc comparisons revealed significant differences only between the groups of 0 minutes and 8 minutes per day. Girls, playing computer or video games an average of 8 minutes per day, had less sexualized perception of romantic relationships than those girls who did not play games.

[everild-blog]

Week 1: running an ANOVA

I would like to go on with the interpretation of the data from the AddHealth data spreadsheet, concerning the amount of time the adolescents spend playing computer or video games and the proportion of sex-related events in their concept of a perfect romantic relationships. You might want to take a look at the previous posts to get some idea of the preceding work.

In fact, my independent (sex, time spent playing video games) and dependent (number and proportion of the sex-related events, the place of the first sexually related event) variables are categorical, so ANOVA is not the most suitable tool to use. 

However, just to have some practice, I performed an ANOVA to answer the following question: is there any difference between male and female teenagers in terms of the time they spend playing video games?

In this case, my categorical variable would be BIO_SEX (you can check out the description of the variables in these posts: http://everild-blog.tumblr.com/post/147185430144/week-2-defining-the-variables and http://everild-blog.tumblr.com/post/147452559879/week-3-data-management), taking only two values: 1 (male) or 2 (female). 

My quantitative variable would be H1DA10, representing the number of hours a respondent spent playing video games per week. This variable takes values in the range of 0-99.

Of course, just like in the previous posts, I decided to include only those repondents who were answering the questions in the absence of a third person (H1IR6=0). Missing data were also managed as following:

IF H1DA10=996 THEN H1DA10=.; IF H1DA10=998 THEN H1DA10=.; IF BIO_SEX=6 THEN BIO_SEX=.;

The null hypothesis was that male and female teenagers spend equal amount of time playing computer or video games.

My code looked like this:

The results of the ANOVA clearly indicate that we should reject the null hypothesis:

Therefore, male teenagers play computer and video games significantly more than female teenagers. As my categorical variable included only two groups, I did not have to perform post-hoc analysis.

[everild-blog]

Week 4: visualizing data

In the previous post I have chosen four variables from the AddHealth data spreadsheet for the analysis:

Biological sex of the respondent (BIO_SEX);The amount of time spent playing computer/video games per week (TIME_GROUP);The proportion of sexually related activities within the first five suggested romantic events (SEX_SUM);The number of the first sexually related event (SEX_FIRST).

My question is whether playing computer games is associated with sexualized romantic behaviour in adolescents.

Here is the code of my program (click for larger version):

It turns out to be that all the variables I’ve chosen are categorical, so I was not able to calculate univariate statistics for them. Here are the univariate graphs. I’ve changed the TIME_GROUP variable a little, so that now it represents the number of minutes spent playing computer/video games a day, which would be more convenient for future bivariate graphs. The categories are “not playing”, “playing less than 15 minutes a day”, “playing 15 to 20 minutes a day”, “playing 20 to 45 minutes a day”, “playing more than 45 minutes a day”.

The distribution of this variable is more or less uniform.

The distribution of this value is unimodal, asymmetrical, skewed-right. The majority of respondents do not play video or computer games.

The distribution of this value is unimodal, asymmetrical, skewed-right. The majority of the respondents do not include any sexually related events in the list of the first five events of their ideal romantic relationship.

The distribution of this variable is bimodal, skewed-left. The majority of the respondents  do not include any sexually related events in the list of the first five events of their ideal romantic relationship; among the rest, the most common place for the first sexually related event was the 4th one.

I wanted to explore the association between the variables TIME_GROUP, SEX_SUM and SEX_FIRST. TIME_GROUP is the explanatory variable, while SEX_SUM and SEX_FIRST are the response variables.

In the first case, the response variable, SEX_SUM, is categorical. It contains 6 categories, therefore, I had to collapse it into 2 categories. To do that, I created a new variable, PROPORTION, which was set to 0, if SEX_SUM was 0, and to 1, if SEX_SUM was bigger than 0.

Here is the resulting bivariate graph:

It turns out to be that about 47% of adolescents who do not play computer games have 1 or more sexually related events on their list, while in any group of teenagers playing computer games only 38-39% have it.

In the second case, the response variable, SEX_FIRST, is also categorical, having 6 categories. To collapse it, I created one more variable, EARLY, which was set to 0, if there were no sexually related events within the first five, and to 1, if there was at least one of them. In this case, the resulting graph will, obviously, be identical to the previous one, therefore I decided to set EARLY to 0, if there were no sexually related events within the first three, and to 1, if there was at least one of them.

There seems to be a general negative association between these two variables, but there is a prominent peak corresponding to the last group of the most “hardcore” gamers, playing more than 45 minutes a day (5 hours a week).

It was also interesting for me to see if there are any differences between male and female respondents. I repeated the previous steps, including only male or only female teenagers in the analysis, and got the following graphs:

As you can see, there is a major discrepancy between male and female gamers of the group playing 15 to 20 minutes a day (about 1.5-2 hours per week). While the male respondents of this category (n=304) seem to be less interested in sexual activities, the female respondents (n=225), in fact, tend to have more sexualized perception of romantic relationships.

[everild-blog]

Week 3: data management

I received several fair comments on the previous post, where I was first introducing my data. In fact, I was totally agree with those notes, so I decided to improve my data and make them easier to understand. Luckily, this week's assignment was just what I needed :)

Here are the screenshots of my code (click the pictures to get high-res versions):

First of all, I decided to get rid of the missing data. I excluded the participants who were answering the questions in the presence of a third person, as it might have biased their answers (variable H1IR6); then I marked the answers "refused to answer", "don't know", "not applicable" and "legitimate skip" as missing data.

The variable representing the number of hours a respondent spent playing video games per week was originally taking values from 0 to 99. I decided to create several categories. Approximately half of respondents was not playing computer games at all. I decided to divide the rest of the answers into three groups, coming up with the following categories:

not playing computer games (i.e., 0 hours per week);

playing 1 hour per week;

playing 2-4 hours per week;

playing more than 4 hours per week.

I used a new variable TIME_GROUP to represent these categories.

The frequency tables of the variables representing the suggested first romantic event (you can get the whole list of answer options here) were too confusing, as well. I decided to manage them the following way:

First of all, I was not interested in particular options; the thing I really wanted to know was whether they were or were not sexually related. Therefore, I created 5 new variables (for the first 5 romantic events suggested) SEX_1, SEX_2, SEX_3, SEX_4, SEX_5 and made them equal to either 0 (if this event was not sexually related - i.e., the value variable was not in the range 12-16) or 1 (if it was). The third option was “missing value”, if the respondent refused to answer or if he/she was out of cards by the moment (legitimate skip).

Here is an example. Let us consider a respondent, who was asked to put the cards in the order in which he/she thought things should first happen in his/her perfect relationship. The respondent puts the cards in the following order:

We would go out together in a group. (H1ID1_1=1)

We would go out together alone. (H1ID1_2=5)

We would kiss. (H1ID1_3=13)

We would have sex. (H1ID1_4=15)

legitimate skip [all cards rejected or refused] (H1ID1_5=., as we've marked this kind of answer as "missing data" earlier)

In this case we assign variables SEX_1 and SEX_2 0, variables SEX_3 and SEX_4 1, variable SEX_5 . (missing data).

The next thing I wanted to know is how many of the romantic events were sexually related. In order to do that, I created a new variable SEX_SUM, which summarized the values of the SEX_# variables or was set to ., if at least one of the SEX_# variables was "missing". The value of SEX_SUM was then divided by 5, so I got a proportion of sexually related events.

In our example, SEX_SUM would be set to "missing", as there was no fifth answer.

However, if the respondent had actually given the fifth answer (for example, "We would get married", which is not considered sexually related), SEX_SUM would be equal to 0.4, representing 40% of sexually related events.

I was also interested in how early the sexual activity aroused. To examine it, I created one more variable, SEX_FIRST, which takes the value equal to the number of the first sexually related event (i.e., of the first SEX_# variable equal to 1) or, if there are none, equal to 6 (which was later encoded as "None of the first 5 romantic events is sexually related").

In our example, SEX_FIRST would be set to "missing data".

In case if the respondent had actually provided 5 answers, SEX_FIRST would be equal to 3.

The last step was to rename the output values to make them more obvious. I performed this procedure for the variable BIO_SEX (representing biological sex of the respondent), TIME_GROUP (representing the number of hours spent playing computer games) and SEX_FIRST (representing the number of the first sexually related romantic event).

Here is the picture of my resulting frequency tables:

You can see that there was an approximately equal amount of male and female teenagers participating (49.79% and 50.21%, respectively). Approximately half of them (47.17%) did not play computer games, and only 17.03% played more than 4 hours a week. There were also 7 respondents refusing or unable to answer.

57.63% of participants were not willing to have any sexually related events among the first 5 events of their perfect romantic relationship, and approximately 95% of the respondents were not willing to have more than one event. However, there was even a respondent was a 100% sexualized perception of the beginning of a perfect relationship. 46 answers were excluded due to the lack of data.

The most common place for the first sexually related event was the 4th one; however, there was a significant disagreement on this point, so the data is distributed more or less evenly.

[everild-blog]

Week 2: running the first program

Here you can get the information on the variables I chose for my research. Briefly, these include biological sex, number of hours spent playing video games per week, first five events that should occur in the perfect romantic relationship according to the respondent.

Here is the code I wrote in order to subset the raw data, to sort it and to create frequency tables:

Here are the frequency tables on the variables I’ve chosen:

A random sample of 4,622 adolescents (2,296 or 49.68% male and 2,326 or 50.32% female) were asked the following question: “How many hours a week do you play video or computer games?”. Of the total number, 47.19% answered “0″ (thus NOT playing computer games). Others chose a number in the range of 1 to 99 hours per week. The majority of them chose 1 to 7 hours (43.25% of the total number). The answers of the rest of the respondents (442 adolescents) were distributed almost uniformely with a prominent peak at 10 hours (125 respondents, 2.70% of the total number).

For the next question, the same students were asked to choose the order in which they thought things should first happen in their perfect relationship. There was an initial set of 17 cards (you can see the list here), and the respondents were free to discard those which seemed not applicable to them.

Of the following number, 3.33% (154 respondents) chose one of the cards describing sexual activities (answers 12-16) to be the first thing to happen. The second event was related to sex in 2.70% (125 respondents). 10.51% (486 respondents) put a card describing some sexual activity in the third place. The fourth event was related to sex in 14.89% (688 respondents), while the fifth one - in 13.41% (620 respondents).

[everild-blog]

Week 2: defining the variables

As I’ve mentioned in my previous post, I decided to conduct my research on whether teenagers who play videogames have more sexualized perception of the ideal romantic relationship or not (or, maybe, there is no such connection at all).

I’m working on the AddHealth data set, which contains the survey results. I’ve chosen the following variables for my analysis:

The number of hours of playing computer/video games per week (H1DA10). The respondents gave the number from 0 to 99; there were also teenagers who refused to answer or said they didn’t know.

The first five things that would possibly happen in course of the ideal romantic relationship (H1ID2_1, H1ID2_2, H1ID2_3, H1ID2_4, H1ID2_5). There is a set of answers, and the respondents were to put them in the order they thought these events would happen. Here is the list of possible answers (excluding the legitimate skip [if the respondent was thinking that none of these would happen in his/her ideal relationship], the refusal and the “don’t know” answer):

We would go out together in a group.

I would meet my partner’s parents.

I would tell other people that we were a couple.

I would see less of my other friends so I could spend more time with my partner.

We would go out together alone.

We would hold hands.

I would give my partner a present.

My partner would give me a present.

I would tell my partner that I loved him or her.

My partner would tell me that he or she loved me.

We would think of ourselves as a couple.

We would talk about contraception or sexually transmitted diseases.

We would kiss.

We would touch each other under our clothing or with no clothes on.

We would have sex.

My partner or I would get pregnant.

We would get married.

The answers 12-16 seem to be describing various sexual activities. My aim is to see whether these answers tend to be linked to higher rate of playing video games. It is also possible that the answers will be different depending on gender, so I decided to include the biological sex (BIO_SEX) as a variable. I also noticed that, in many cases, there was a third person present and/or interrupting the survey, so I decided to include only those repondents, who were answering the questions in the absence of a third person (H1IR6=0).

[everild-blog]

Week 1: Choosing the subject, building the hypothesis

Hey there, Courserians! :) My name is Alina, I’m a postgrad student from Russia, studying cell biology. I enjoy playing video games in my spare time (though there’s less and less of it left). I’ve been playing since I was 6, so it is literally a huge part of my life.

I was looking through the AddHealth codebook and saw that there is a large section devoted to ideal romantic relationship. I was intrigued, so I looked through this section and found an elaborate questionnaire, containing A LOT of information on the concept of a perfect relationship of adolescents.

I realized that, in fact, I would like to look for some patterns in the perception of romantic love by teenagers. It came to my mind that there is a number of papers deling with violence and video games; however, I haven’t seen any on the connection between love and video games. Yet, many popular games give the player a chance to conduct romantic relationships, to make certain choices - and, at the same time, limit those choices, establish certain sequences of actions, thus giving the players some prefabricated patterns of romantic behaviour. So, I decided to take up the possible correlation between gaming experience and perception of romantic relationship as a subject.

I’m going to take the number of times and the duration of video game playing per week and the probable components of ideal romantic relationship (their quality and sequence) as my variables.

I made up my personal codebook and, for now, included Section 2 and Section 21 into it (Daily Activities and Ideal Romantic Relationship, respectively).

Having chosen the topic of my study, I performed a literature review on the subject (video games & romantic relationships), using Google Scholar. The information on this topic turned out to be really sparse (I mean scientific papers, not just magazine articles providing point-of-view with no real evidence). Still, there are certain studies related to the subject.

1. There might be a connection between our in-game and real-world behaviour.For example, the article by Greitemeyer and Osswald (2011,http://www.tandfonline.com/doi/abs/10.1080/00224540903365588) suggests thatexposure to prosocial video games increases the accessibility of prosocial thoughts; furthermore, the study by Greitemeyer, Osswald and Brauer (2010,http://dx.doi.org/10.1037/a0020194) says that exposure to prosocial video games is positively related to prosocial behaviour, increasing empathy and decreasing pleasure at another’s misfortune.

2. Video games may influence sexual behaviour and sexual beliefs. The research by Dill and Thill (2007, http://link.springer.com/article/10.1007/s11199-007-9278-1) shows that video games provide certain gender stereotypes, accepted by gamers. Male game characters are represented as aggressive, while female ones - as hypersexualized, often combining incompatible features (e.g., both slim and having large breasts). At the same time, the article by Bleakley et al. (2008,http://www.tandfonline.com/doi/abs/10.1080/15213260802491986) proves thatexposure to sexual content in media leads to progress in sexual behaviour in adolescents (deep kissing, touching each other, different forms of sex). The review by Manganello (2008, http://tva.sagepub.com/content/9/1/3.short) suggests a potential pathway for dating violence to be learned from observing media.

Therefore, based on the literature review, I developed a hypothesis for my future work:extensive gaming experience influences the perception of ideal romantic relationship. It is possible that the concept of perfect relationship might be more sexualized in the adolescents spending more time playing video games.