Re: Which Statistical test to use?!

Posted by Art Kendall on Dec 18, 2012; 1:53pm
URL: http://spssx-discussion.165.s1.nabble.com/Which-Statistical-test-to-use-tp5717005p5717018.html

multiple regression is a special case of GLM. but doing the GLM with a anova with repeated measures aaproach would use more of the information available.

Also if you have missing data in some of the repeats, you might need to move to multiple level modelling.
On 12/18/2012 4:39 AM, Saeed Rezaei wrote:

Thanks John for your reply. I have already done the descriptive statistics on the items and their related likert scale (providing the frequency and percentage).
A friend of mine said I better use multiple regression but change the research question to prediction. I am still a bit confused!

Thanks for your help

On Tue, Dec 18, 2012 at 8:27 PM, John F Hall <[hidden email]> wrote:

Have you tried running a simple crosstabs of your demographics against your scales?

Not exactly clear from your email what you’ve got in the way of data, but from your data editor:

File > New > Syntax

. . then write in: [modify to your own variable names]

CROSSTABS gender, age, level by scale1 to scale6

/cel cou row /sta chi.

. . Run > all

It looks like you need to think of analysis of variance (ANOVA) to compare more than two groups on each of your dependent variables, but others on the list are better statisticians than me. Get back to me if you need further help.

John F Hall (Mr)

[retired academic survey researcher]

Email: [hidden email]

Website: www.surveyresearch.weebly.com

From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Saeed Rezaei
Sent: 18 December 2012 09:22
To: [hidden email]

Subject: Re: Which Statistical test to use?!

Thanks for your reply.
Actually I have never run decision tree for my data analysis!
I hope I can work out an easier way!
Thanks and I will be looking forward to hearing more from the group and you.

On Tue, Dec 18, 2012 at 7:05 PM, DorraJ Oet <[hidden email]> wrote:

Hi Saeed,

It is nice to know that each item is likert scale.

The reason why I suggested a decision tree because a Dean of a University once mentioned that the decision tree take care of interaction. Also, it might be meaningful to find out which of the variables, gender, age and proficiency level is important to

1. attachment to persian language
2. pronunciation attitude
3. language and social status
4. L1 use and exposure in the society
5. language knowledge
6. script

If I am doing the analysis, I will build decicison trees for all the 6 items using them as dependent variables and gender, age and proficiency level will be the independent variables. It might be interesting to know which of these 3 variables are important to the 6 variables above.

One more thing about Decision tree is that is can actually "group" the age for me. The age range of a respondent being attached to Persian Language might be different from that who has a high language knowledge. I can also do profiling along the way with a decision tree.


Warmest Regards
Dorraj

---

Date: Tue, 18 Dec 2012 18:45:51 +1100
Subject: Re: Which Statistical test to use?!
From: [hidden email]
To: [hidden email]
CC: [hidden email]

I have already run EFA and CFA on the questionnaire. Actually I have the Likert scale from 1 to 6 for each of the items in the questionnaire, and the gender is male and female as expected, the age group is divided into four 11-15, 16-20, 21-25, 26+, and the language proficiency is also divided into six including basic, elementary, pre-inter, inter, high inter, and advanced.
My first question is: Should I check the relationship between each subscale and each of the above variables? i mean does it sound plausible? And if yes, it becomes very complicated for data analysis. What do you suggest?
Thanks

On Tue, Dec 18, 2012 at 6:39 PM, DorraJ Oet <[hidden email]> wrote:

Dorraj

---

From: [hidden email]
To: [hidden email]; [hidden email]
Subject: RE: Which Statistical test to use?!
Date: Tue, 18 Dec 2012 07:39:55 +0000

Hi Saeed,

This seems to be more of a modeling than a Statistical testing to see if there is a relationship.

I am going to try to suggest based on what I had interpret from your email.

Looking at if there is a relationship between Iranian English language learners’ attachment to Persian

1. gender
2. age
3. proficiency level

My dependent variable will be Attachment to Persian. Assuming that attachment is scored 1 - 5.

My independent variables will be gender, age and proficiency level.

I will use a decision tree model to test which are the independent variables that are important to the variable Attachment to Persian.

If all 3 of the independent variables are important, the decision tree will allow you to see

For a respondent which is

1. Male
2. 30-51 years old
3. Proficiency level 5

how attached is he to Persian.

Warmest Regards
Dorraj Oet

---

Date: Tue, 18 Dec 2012 18:01:17 +1100
From: [hidden email]
Subject: Which Statistical test to use?!
To: [hidden email]

Hi all,
I am dubious about which statistical test to use for my research questions. I would be glad to receive your comments. My research question is:

Is there any relationship between the Iranian English language learners' language identity and their age, gender, and language proficiency level.
The language identity is measured through a validated questionnaire with 19 items. And the language identity included six subscales in the questionnaire namely:
1. attachment to persian language
2. pronunciation attitude
3. language and social status
4. L1 use and exposure in the society
5. language knowledge
6. script

Each of the above subscales had 3 to 4 items.

Now the first research question will be transformed into multiple research questions including:

1. There is no relationship between Iranian English language learners’ attachment to Persian and their

· gender,

· age, and

· proficiency level

2. There is no relationship between Iranian English language learners’ pronunciation attitude and their

· gender,

· age, and

· proficiency level

3. There is no relationship between Iranian English language learners’ language and social status and their

· gender,

· age, and

· proficiency level

4. There is no relationship between Iranian English language learners’ L1 use/exposure in the society and their

· gender,

· age, and

· proficiency level

5. There is no relationship between Iranian English language learners’ language knowledge and their

· gender,

· age, and

· proficiency level

6. There is no relationship between Iranian English language learners’ script/alphabet and their

· gender,

· age, and

· proficiency level

Now....what do you think would be the best statistical test? Do you think I should use MANOVA? Do not forget that I am after "relation" not "effect". How should I enter the data for each subscale.
Thanks in advance to all of you for your kind attention.

Saeed

===================== 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