Which test do I have to use?

Moderators: statman, Analyst Techy, andris, Fierce, GerineL, Smash

Lisan19hotmail.c92
Posts: 5
Joined: Fri Apr 10, 2015 8:48 am

Which test do I have to use?

Postby Lisan19hotmail.c92 » Fri Apr 10, 2015 9:12 am

Hi there,

Thanks a lot for reading this!
I'm struggling with SPSS for a long time now, so I'm happy that I found this forum.
I'm writing a master thesis about children from different ethnic groups and friendships. I have two research questions and I don't know which SPSS test I have to use. I hope someone can help me with 1 or both of my questions.

RQ1: Do children from a particulair ethnic group have more friendships with children from the same ethnic group or does the ethnic group not matter in choices of friendships?
For this RQ, I used a questionnaire with 250 learners from 7 different ethnic groups. In the questionnaire I asked who their friends from the class where and I noted the ethnic group from each friend.

RQ2: Do children form various ethnic groups value friendships different?
For this RQ, I used a questionnaire with 250 learners from 7 different ethnic groups. The questionnaire consisted of 40 statements. Each statement could be answered with: not nice, nice, very nice. The 40 statements consisted of three underlying factors (appeared from the factor analysis).

Which SPSS test can I use best for these RQs? The different ethnic groups are not equal. That's definitely a problem in RQ1. I already added a percentage of how big the chance is to came from a certain ethnic group, but I still don't know which test I have to use.

Thanks in advance
GerineL
Moderator
Posts: 1477
Joined: Tue Jun 10, 2008 4:50 pm

Re: Which test do I have to use?

Postby GerineL » Fri Apr 10, 2015 2:04 pm

Lisan19hotmail.c92 wrote: RQ1: Do children from a particulair ethnic group have more friendships with children from the same ethnic group or does the ethnic group not matter in choices of friendships?
For this RQ, I used a questionnaire with 250 learners from 7 different ethnic groups. In the questionnaire I asked who their friends from the class where and I noted the ethnic group from each friend.
This sounds like a question you might want to answer using social network analyses programs such as sienna.
If you do want to answer it using spss, in a less sophisticated manner then of course, you could compute the number of friends (or percentage) of the same ethnic group vs. other ethnic group, and compare those by ethnicity.
For this RQ, I used a questionnaire with 250 learners from 7 different ethnic groups. The questionnaire consisted of 40 statements. Each statement could be answered with: not nice, nice, very nice. The 40 statements consisted of three underlying factors (appeared from the factor analysis).
if your groups are much different in size, you can use a kruskall wallis test to see if there are differences at all, and then use mann-whitney u test for paired comparisons.
same goes for the question above by the way.
Lisan19hotmail.c92
Posts: 5
Joined: Fri Apr 10, 2015 8:48 am

Re: Which test do I have to use?

Postby Lisan19hotmail.c92 » Sat Apr 11, 2015 12:08 pm

Thanks for answering my questions! I have two more questions:
- Is it not better to do a Regression for RQ1?
- I don't see where I can do a Mann Whitney U test for more than 2 groups (I have 7 different ethnic groups). Do you know where I can find it or isn't that possible?
GerineL
Moderator
Posts: 1477
Joined: Tue Jun 10, 2008 4:50 pm

Re: Which test do I have to use?

Postby GerineL » Mon Apr 13, 2015 7:43 am

Lisan19hotmail.c92 wrote:Thanks for answering my questions! I have two more questions:
- Is it not better to do a Regression for RQ1?
There is lots of literature about this question (referred to as selection in literature).
Of course you can use less sophisticated techniques, but the field has evolved and nowadays these types of techniques are used more often. Techniques such as sienna take into account more dynamics of friendship choices etc.
I am not sure if regression is the way to go, it depends a bit on how your data are coded, can you say something about that?
- I don't see where I can do a Mann Whitney U test for more than 2 groups (I have 7 different ethnic groups). Do you know where I can find it or isn't that possible?
if you have 7 groups, you need planned comparisons. For instance, I can imagine that you might not be interested in all differences between all groups. You do pairwise comparisons between all your groups of interest. Just a note, if you would use anova with post-hoc comparisons, you would also have 21 comparisons.
Lisan19hotmail.c92
Posts: 5
Joined: Fri Apr 10, 2015 8:48 am

Re: Which test do I have to use?

Postby Lisan19hotmail.c92 » Mon Apr 13, 2015 8:42 am

There is lots of literature about this question (referred to as selection in literature).
Of course you can use less sophisticated techniques, but the field has evolved and nowadays these types of techniques are used more often. Techniques such as sienna take into account more dynamics of friendship choices etc.
I am not sure if regression is the way to go, it depends a bit on how your data are coded, can you say something about that?


Each participant could choose as much friends as they want to. I knew the ethnic background of each friend and noted that in SPSS.
GerineL
Moderator
Posts: 1477
Joined: Tue Jun 10, 2008 4:50 pm

Re: Which test do I have to use?

Postby GerineL » Mon Apr 13, 2015 10:53 am

ok so what you could do is recode and sum the percentage of "same ethnicity", and run a kruskall wallis test to see whether this differs per ethnicity.
Lisan19hotmail.c92
Posts: 5
Joined: Fri Apr 10, 2015 8:48 am

Re: Which test do I have to use?

Postby Lisan19hotmail.c92 » Sat Apr 18, 2015 9:36 am

Thanks for answering my questions. I have another question about my second RQ.
RQ2: Do children form various ethnic groups value friendships different?
For this RQ, I used a questionnaire with 250 learners from 7 different ethnic groups. The questionnaire consisted of 40 statements. Each statement could be answered with: not nice, nice, very nice. The 40 statements consisted of three underlying factors (appeared from the factor analysis). The three underlying factors are: 1. Recreation and Advice, 2. Intimacy, 3. Conflicts
The Cronbach's alpha from factor 1 and 2 are high (.9). But the Cronbach's alpha of factor 3 (conflicts) is low (.399). So I tried to recode the items from this factor. But when I do this, the Cronbach's alpha is very low (.022)! The items in this factor are like this:
'Me and my friend are fighting a lot'. Although it's a negative item, it is part of friendship. So is recoding the solution or do I have to do something else?
GerineL
Moderator
Posts: 1477
Joined: Tue Jun 10, 2008 4:50 pm

Re: Which test do I have to use?

Postby GerineL » Mon Apr 20, 2015 9:09 am

If you compute a scale, it is important that items are recoded if they are framed in opposite direction.
for instance:

"Me and my friend are fighting a lot" (scoring high on this item means having high conflict)
vs.
"My friends and I never disagree" (scoring high on this item means having low conflict).

So if you would compute a total scale score, you would probably recode the negatively formulated items (which might be the conflict items).
If you compute a separate subscale for conflict, you only have to recode those items that are framed in the opposite direction, as indicated above.

If the scale is an existing scale, try to find a manual to see which items need recoding etc.
sarwanov
Posts: 1
Joined: Sun May 03, 2015 6:28 am

Re: Which test do I have to use?

Postby sarwanov » Sun May 03, 2015 6:31 am

It depends on tests.choose most suitable ones
http://www.soran.edu.iq

Who is online

Users browsing this forum: No registered users and 1 guest

cron