Likert Scale analysis

[Charlotte1]

Hi I am new and need help asap!

right, ive done a questionnaire and it has two 5 point likert scale questions within it, so far ive done factor analysis.

Now I want to see how responses to these questions differ amongst gender and profession....what statistical analysis method do i need to use? and can it be done in SPSS?

thanks for your help

[Dr.D]

Well you can use independent samples t-test to examine gender differences on responses to Likert scale items, and if profession has three or more categories or groups of persons, then you can use a one-way ANOVA.

Go to menubar in SPSS and Analyse, Compare Means, to find each test. For t-test, put gender in grouping variable box, and Likert items in other bigger box (Test Variable).

For ANOVA, put profession in factor box, and Likert items in dependent box. Go to options and select descriptive statistics. Then go to post-hoc and select Boneferroni test.

You can look at the means of each group to determine how strongly one group feels in relation to another group on a particular issue. Then you look at Sig. level (which is the p value) to determine whether the groups differ significantly in their views based on the p-value being less than .05.

[Charlotte1]

Hi

Thank you soo much for replying, the profession question is basically a yes or no question with rgeards to whether the repsondent was a student or not, so could i use the t-test again for that??

Someone has mentioned doing a mann whitney test....would that be suitable here at all?

Also within the likert scale items i want to determine which factors are the most important in determining a customers overall attitude towards a letting agent, and i believe linear regression is suitbale for this, is this correct? if it is....how do i conduct this on spss?

Thanks again, your a lifesaver

[Dr.D]

Hi

Yes, if profession has two categories (yes or no), you can do a t-test.

Well, a Mann-Whitney test is actually non-parametric t-test (that is it does the same thing as the t-test but on ordinal or non-normal data). You see people argue sometimes that if you have Likert scales, they should be treated as ordinal variables and Mann-Whitney tests are thus suitable. However, you examine mean ranks or medians between groups. But in the literature, you will find that people use normal independent t-tests with Likert items, so it should be ok.

So if you are looking at overall attitude (which I assume are the individual item scores combined as a composite/summed score) based on multiple factors, you need to do multiple linear regression.

In SPSS, go to Analyse, Regression, Linear.
In the dialogue box, put in the factors as independent variables, and the composite score or summed attitude score you put in dependent variable. But you would need to inform me of the factors, and what constitutes the overall attitude, so I will have sufficient information to make a good decision.

[Charlotte1]

hi

well basically the question is made up of 19 statements which a 5 point scale which ranges from agree to disagree. basically i want to see which are the most important when deciding on a letting agent to use... so not sure if linear regression is at all appropriate??

I will conduct the t-test, but are there any tips on how to interpret the information, i.e. putting it into writing?

[Dr.D]

hi

well basically the question is made up of 19 statements which a 5 point scale which ranges from agree to disagree. basically i want to see which are the most important when deciding on a letting agent to use... so not sure if linear regression is at all appropriate??

I will conduct the t-test, but are there any tips on how to interpret the information, i.e. putting it into writing?

See this website

http://www.essex.ac.uk/psychology/ps...p3write-up.pdf


This is also a very good site (go here first)

http://wwwstage.valpo.edu/other/dabook/ch9/c9-3.htm

In terms of your regression issue, what are your independent and dependent variables (I am unclear about this part).

[Charlotte1]

er i guess this is where i really get stuck...im not sure how to determine independent and dependent variables from likert scale data

[Dr.D]

ok tell me what are your research questions/objectives/hypotheses?

I need to see the instrument or data, and I can tell you what you need to do as I am running on little information on the topic (lol)

We will get this organised..dont worry my love

[Charlotte1]

my research objective is :to identify factors important to consumers when selecting a letting agent

there are 19 statements rgearding letting agent...such as location of company, quality of houses, staff advice etc and then the respondent had to say how important that factor was in determining whether they would use that letting agent or not. the scale was 1 being very important through to 5 which was not important at all....

For each scale (1-5) there is a % for each. so like 45% said the location of the company was fairly improtant and so forth.

[nicky nguyen]

Dear
This is the first time I worked on thesis but I faced one confused question. Why most people used Likert scale for survey from strongly disagree to strongly agree. What is the difference between strong not important to important.

2. How to know exactly how many sampling that we will collect.
eg: There are 100 hotels in the area which I will collect. But i can't go all 100 hotels. But I want my data is reliable to be input on SPSS and provide a useful results. How to do it and what is formulation.
How many sampling that we will forecast not valid after collection. Thanks and appreciate for your speedy reply. Thanks, Nicky

[viwel]

Hi,
I have a question about this scale:
We have elaborate questions about a new methodology of teaching
The questionnary are formulate following a Likert scale.

Ex.: This new methodology is appreciate by the residents:
Desagree................... Agree
-----1 .......2.....3.....4........5
We have send this questionnary to several profs.

Following you, how can I analyse these answears ???
1/
For each questions, I calculate the mean with a DEV.
Ex.: This new methodology is appreciate by the residents
Mean: 3,5 +/- 0,8

2/
For each question, I calculate the pourcentage of answear recevied:
Ex.: This new methodology is appreciate by the residents:
60% answear 3
30% answear 4
10% answear 5
Which test must I do to see if there is a difference between answears ???
A Chi-2 ???

3/
Other methodes ???

Regards

[viwel]

No idea ???

[TheEcologist]

No idea ???

You’re intuition on doing a chi-squared seems right. You can test the actual number of people that answered a particular value in each question and compare it to the expected or “random”, where each value in each question has an equal chance of appearing. Just don’t use percentages in the chi-squared but the number of people who answer it.

cheers,

[viwel]

Regards, but I d'ont understand why I don’t use percentages in the chi-squared but the number of people who answer it ?

[TheEcologist]

Regards, but I d'ont understand why I don’t use percentages in the chi-squared but the number of people who answer it ?

Chi-square requires that you use numerical values, not percentages or ratios.
Just look at the formula for calculating the chi-squared statistic.

Sum of all: (observed-expected)^2 / expected

Example:

If you have 45 respondents to a yes/no poll

10 say yes
35 say no

You expected a 50:50 ratio

The chi-squared using individual responses is:

(10-22.5)^2/22.5 + (35-22.5)^2/22.5 = 13.9

The chi-squared value using percentages is:

(22.2%-50%)^2/50% + (77.8%-50%)^2/50% = 30.9

Clearly not the same…