Re: Help with Dissertation SPSS

Posted by Mike on Aug 12, 2011; 10:10pm
URL: http://spssx-discussion.165.s1.nabble.com/Help-with-Dissertation-SPSS-tp4693273p4694753.html

To add to Bruce's comments:

It seems to me that there is a confusion about what your independent
variable is.  If one goes back to the OP's original post, one finds:

>My study is the effects of video games on prosocial behavior.
>
>2 DV = prosocial behavior and aggression - measured with a 38 item
>questionnaire (15 Aggressive/23 Prosocial)
>
>IV = I used a video game questionnaire where participants are asked to
>rate the frequency of gameplay they have had in the past month (1 = no days,
>2 = 1-5 days, 3 = 6-10 days, 4 = 11-20 days, 5 = more than 20 days). This is
>done for 8 games (RPG, adventure, strategy, platform, action and fight,
>first person shooter, sports, racers). These games are divided into 3
>groups, violent, non-violent and neutral.

Is your independent variable (a) number of days spent playing video
games or (b) type of video game playes (i.e., violent, non-violent,
and neutral)?

You seem to indicate that (b) is the answer given what you said earlier:

>The proposed method which was approved was using a One-way between
>subjects ANOVA with 1 DV (Prosocial) and 1 IV Video games (Violent,
>non-violent, neutral).

If this is correct, what is the role for "number of days in past month playing
games"?  I would assume that people who played violent video games most
often in the past month would go into the "violent video game" group,
those who spent most of the past month playing "nonviolent video games"
would go into the "nonviolent video game group", and those who spent
most of the time playing neutral games were put into the "neutral video
game group".

Is my interpretation correct?  If so, then go do a one-way ANOVA
with type of video game (played most often in the past month) as the
between subject factor and the two scores from your questionnaire
(i.e., measure of (i) prosocial behavior and (ii) aggressive behavior).

If my interpretation is incorrect, then identify how so.

I think that the focus on number of days a game is played is misplaced,
if my assessment above is right.

In any case, even with three level "Type of Video Game" between-subject
factor, one could also do a regression analysis where the three groups
are coded into two predictors, following Bruce's suggestion.

-Mike Palij
New York University
[hidden email]



----- Original Message -----
From: "Bruce Weaver" <[hidden email]>
To: <[hidden email]>
Sent: Friday, August 12, 2011 5:43 PM
Subject: Re: Help with Dissertation SPSS

> Hello Muralli.  I was just reacting to Sonia's statement that one cannot have
> categorical explanatory variables in regression models, and did not attempt
> or intend to address all of the questions in your original post.  It seemed
> to me when I read your post that you were asking questions that cannot
> easily be addressed with the exchange of a few messages on this list.
> (Others may disagree, of course.)  Here are some questions for you.
>
> 1. What discipline are you in?  (I would guess psychology from the content.)
> And what degree are you working on (i.e., undergrad, masters, PhD)?
>
> 2. How many stats courses have you taken, and at what level?  Were ANOVA,
> regression & linear models covered?
>
> 3. Have you asked your supervisor (or other committee members) for guidance?
>
> 4. Have you asked your stats instructor(s) for help?
>
> 5. Is there a statistical consulting service at your
> school/college/university?
>
> What I am driving at is that I think you would benefit from meeting with
> someone local.
>
> HTH.
>
>
> Muralli wrote:
>>
>> I'm sorry Bruce but I don't understand your explanation, the research
>> was meant to be an easy one but it would appear that the surveys i
>> used has made it complicated. My aim is to measure the effects of
>> video games on prosocial behavior. My hypothesis are video games will
>> increase prosocial behavior, and video games will not increase violent
>> behavior. In my proposal that was approved, i stated that I'm using 1
>> IV which is types of games (Violent, non-violent, neutral) and 1 DV
>> prosocial behavior and that the design will be a one-way between
>> subjects ANOVA. Is there a way to still do it like the way i proposed?
>>
>> P.S. I'm not sure how to send a reply, I'm sending one to all your
>> emails and also to SPSSX-L hope you guys receive it.
>>
>> On 8/13/11, Bruce Weaver &lt;[hidden email]&gt; wrote:
>>> Hi Sonia.  The "days" variable is:  1 = no days, 2 = 1-5 days, 3 = 6-10
>>> days,
>>> 4 = 11-20 days, 5 = more than 20 days.  I would describe that as ordered
>>> categories, not nominal.  But in either case, categorical variable like
>>> this
>>> certainly *can* be used in regression models, but one has to compute k-1
>>> indicator variables to do so (where k = the number of categories).  A
>>> DO-REPEAT is a nice way to do that.  E.g.,
>>>
>>> do repeat f = f1 to f5 / # = 1 to 5.
>>> - compute f = (days EQ #).
>>> end repeat.
>>>
>>> Then use 4 of the 5 f-variables to code for frequency of play.  The
>>> omitted
>>> 5th f-variable will be the reference category for the t-tests you see in
>>> the
>>> table of regression coefficients.  Using 4 of the 5 indicators for "days"
>>> this way in REGRESSION gives exactly the same result you get by treating
>>> the
>>> original "days" variable (with 5 categories) as a fixed factor with
>>> UNIANOVA.  (Make up an example and give it a try.)
>>>
>>> p.s. - I'll save Jon the trouble of adding that there is a Python-based
>>> method for computing indicator variables too.  Most of the people I know
>>> locally do not have Python installed, however, and I'll bet many of them
>>> have never even heard of it.  So I like sticking to native SPSS code when
>>> possible.  ;-)
>>>
>>> HTH.
>>>
>>>
>>>
>>> Sonia Brandon-2 wrote:
>>>>
>>>> Your IV regarding the number of days is most definitely nominal, so you
>>>> shouldn't use regression, especially on something for a dissertation.
>>>> If
>>>> you had asked for the actual number of days, it would be ratio, and then
>>>> regression would be more appropriate.  On your DVs, what are the
>>>> response
>>>> values?  Is it a true scale or a Likert scale - or is it yes/no
>>>> response?
>>>> Just guessing, it sounds like you may have multiple interval DVs, and
>>>> with
>>>> Categorical IVs (type of game and frequency category), you may want to
>>>> take a look at MANOVA as the appropriate technique.
>>>>
>>>>
>>>>
>>>>
>>>> Sonia Brandon, Ph.D.
>>>> Director, Institutional Research
>>>> Colorado Mesa University
>>>> 1100 North Avenue
>>>> Grand Junction, CO  81501
>>>> phone: [phone]
>>>> fax:  [phone]
>>>> email: [hidden email]
>>>>>>> Garry Gelade &lt;[hidden email]&gt; 8/12/2011 11:46 AM
>>>>>>> >>>
>>>> How many subjects do you have? And what is/are your research questions?
>>>> Eg
>>>> Frequency of playing any type of game affects aggressive & prosocial
>>>> behaviours? Or Frequency of playing violent games affects these
>>>> behaviours?
>>>> Or both?
>>>>
>>>> Sticking with the regression for the moment, if I have understood your
>>>> data,
>>>> in the first case, your IV could be the total frequency score for all
>>>> games.
>>>> In the second case you could total the frequency ratings for each type
>>>> of
>>>> game, giving you 3 IVs.
>>>>
>>>>
>>>> Garry
>>>>
>>>>
>>>> -----Original Message-----
>>>> From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
>>>> Muralli
>>>> Sent: 12 August 2011 14:17
>>>> To: [hidden email]
>>>> Subject: Help with Dissertation SPSS
>>>>
>>>> My study is the effects of video games on prosocial behavior.
>>>>
>>>> 2 DV = prosocial behavior and aggression - measured with a 38 item
>>>> questionnaire (15 Aggressive/23 Prosocial)
>>>> IV = I used a video game questionnaire where participants are asked to
>>>> rate
>>>> the frequency of gameplay they have had in the past month (1 = no days,
>>>> 2
>>>> =
>>>> 1-5 days, 3 = 6-10 days, 4 = 11-20 days, 5 = more than 20 days). This is
>>>> done for 8 games (RPG, adventure, strategy, platform, action and fight,
>>>> first person shooter, sports, racers). These games are divided into 3
>>>> groups, violent, non-violent and neutral.
>>>>
>>>> I tried to do a multiple regression since someone told me that my IVs
>>>> are
>>>> not categorical but continuous. I compiled the scores for my DV and
>>>> computed
>>>> with target variable splitting up the aggressive questions and the
>>>> prosocial
>>>> questions (Q1 + Q3 + Q5, etc). Thus, the score comes up to 80, 103, 76
>>>> etc.
>>>>
>>>> Then I used each game as an IV but since they only had to rate the
>>>> frequency, the data for each IV was 6, 3, 5, etc.
>>>>
>>>> Clearly, this led to a non-significant result and then I understood why.
>>>>
>>>> I have to idea how to use my IV and i'm not sure what test to use, I
>>>> hope
>>>> the way I have totaled my DVs were right.
>>>>
>>>> The proposed method which was approved was using a One-way between
>>>> subjects
>>>> ANOVA with 1 DV (Prosocial) and 1 IV Video games (Violent, non-violent,
>>>> neutral).
>>>>
>>>> I don't see how to do this because the prosocial survey measures two
>>>> things
>>>> as mentioned earlier and the video games survey is not categorical. At
>>>> least
>>>> that's what i understand from it
>>>>

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD