How to concluded surveyed data to result

x509

New Member
#1
I am working on a survey where I have to prove/diprove my hypothesis. I am seeking this help/guidance since I am getting hard time to related my "multiple-choice" answers in the survey to conclude/related to my hypothesis

I have 20 questions and each question have 3-4 answers (5 sometime). Lets say if i have a question that

how much time do you spend on games in a week
1. 8 hours
2. 6 hours
3. 4 hours
4. 2 hours

For sake of understanding, lets say my hypothesis says that people of age less than 25 spend more time on games compared to older people. Now if a person has replied option 1 to my this question, how can i conclude it towards my hupothesis ?

What i was thinking that based on number of options in question, I assign score to each option i.e., in this case option 1 would have 100, option 2 would have 75, 50 and 25. By this the more someone is close to my hypthesis will have higher score which will result in ANOVA that i will do at the end ...


Am i on the right track ? Any tips/suggestions/guidlines ?
 

x509

New Member
#2
72 views and still no help ...

I am using ANOVA and I am not even sure if i am required to use it ...

Questions like

1. Do you feel aggressive when you lose a game
Ans a: Yes (value 1)
Ans b: No (value 2)
Ans c: sometimes (value 3)

2. Do you get obsessed with a game if you loose couple of games in a row
Ans a: Yes, I have to defeat him once (value 1)
Ans b: no (value 2)
Anc c: sometimes (value 3)
É
now these 2 variables define the aggression combined for each specific age group. The question is, how would i relate these answers to my hypothesis ? If i do ANOVA, I can tell if one group is significant different with another or not. If they are, then these 2 variables doesn't define the level of aggression but if they aren't then it means these 2 questions relate to the each other (correct me if i am wrong here). That being said, how does these 2 questions actually relate to my hypothesis. do i have to do some co-relation ?
 

Karabiner

TS Contributor
#3
If you want to comapre old versus young with regard to ordinal data (like 2-4-
6-8 hours), then use the Mann-Whitney U test.

If you want to compare old versus young with regard to categorical variables,
like that with yes/no/sometimes, then use Chi². If you assume that no-sometimes-yes
(coded e.g. 1-2-3) is ordinal, then you can use U-test here also.

If you want to know whether 2 ordinal variables here are correlated, then
you could use the Spearman correlation coefficient rho.

I addition, I can't see where to use an ANOVA here.

With kind regards

K.
 

x509

New Member
#4
If you want to comapre old versus young with regard to ordinal data (like 2-4-
6-8 hours), then use the Mann-Whitney U test.

If you want to compare old versus young with regard to categorical variables,
like that with yes/no/sometimes, then use Chi². If you assume that no-sometimes-yes
(coded e.g. 1-2-3) is ordinal, then you can use U-test here also.

If you want to know whether 2 ordinal variables here are correlated, then
you could use the Spearman correlation coefficient rho.

I addition, I can't see where to use an ANOVA here.

With kind regards

K.
First of all, thanks a lot for guiding. I have been freaking out since my deadlines are approaching.

Few more questions.

I have 2 independent variables.
1. How much time one spend on gaming (less than hour, 1-4 hours, 4-6 hours, more than that)
2. Age group (younger than 25, 25-35, older than 35)

There are certain group of questions which i want to test against age group and certain groups of question which I want to test against hours. As I mentioned in my previous example, do i have to apply chi2 on each question w.r.t to the independent variable ? i.e., as per my previous example I have 2 questions which are related to no. of hours someone would spend on games. So would i be doing Chi2 on each question w.r.t hours spent ?

Q: I mean i have aound 25 question, so would i be doing Chi2 for each question and writing in survey ?

I just googled and found this article (I am using SPSS) ... I believe this is what I am looking for ... https://statistics.laerd.com/spss-t...est-for-association-using-spss-statistics.php


A stupid question, is there a way I can use ANOVA ?
 
Last edited:

hlsmith

Less is more. Stay pure. Stay poor.
#5
Designing surveys is an art and a whole field. Pilot test the survey many times and talk to the people who take it. Why not have your example question constructed so the respondent can type in the time in hours. Otherwise, if I play it 100 hours I have to select 8 hours or skip it. Lots and lots of nuances to think about.