Likert Scales

#1
I am currently doing my research for my dissertation on the under-representation of women in the construction industry.

My aim is to show that young people are steered toward traditional subjects which in turn will close doors to careers such as construction for women.

For my questionnaire I have many different kinds of questions, but mainly likert scale type which have ratings from 1 to 10 1 being strongly agree to 10 being strongly disagree. I am at my wits end as to how to analyse these questions, I have looked at all ways and can't seem to find the right one, they are all a little confusing.

My initial thought was standard deviation, however I don't understand how this works, I understand its the variance from the mean and most data will be within 1 standard deviation ie +/- 68% but what does this tell me about the data? Another area I though was quartile range and interquartile range, this would tell me if the question answered gave a reading within bottom 25% or top 75% however I have read in books this has to be done by using frequecy which I can't use in this instance.

Any help and guidance would be appreciated, I don't want my dissertation to fail on the stats analysis part.

Please HELP!!!!!

:shakehead
 
Last edited:

JohnM

TS Contributor
#2
There are tons of ways you can analyze Likert scale data, and it all depends on your research question(s) and exactly what you are trying to measure.

If you could post a couple of example questions and what the scale looks like (i.e., what does a "1" represent and what does a "10" represent), along with your specific research question, I can point you in the right direction.
 
#3
Hi thank you for replying below are the questions related to the likert scale


Question 1
1. The subject choices made at school will have an impact on the type of occupation/career that can be pursued

1 Strongly Agree
2
3
4
5
6
7
8
9
10 Strongly Disagree

This question is trying to prove that children at subject choice age aren't aware of the consequences ie in that they don't choose science subjects and don't realise that this could affect their careers in later life, hence most will probably say disagree

There is no scale in the middle (my tutor told me the 1 - 10 scale was better to analyse) so when analysing I will fill these in with whats appropriate

Question 2

Career choice is important at this stage in my life

Same scale 1 - 10

This question is linked to question 1 in that they think at an early age career choice isn't important but could be impacted on from their subject choice, again the probability is they will say disagree

Question 3

I am aware of initiatives to encourage young people into the construction industry

Same Scale 1 - 10

This question is to identify that all initiatives that England have at the moment to encourage young people into the construction industry aren't working, so I need to find out if young people are aware of any.

Thank you
 

JohnM

TS Contributor
#4
I'm not really in favor of using Likert scales that present more than 5 or 7 choices - I don't think that many people can discriminate how they feel (attitude) with that much precision, and the respondents on these types of larger scales tend not to use the entire scale - you end up with most responses on one end of the scale, or with most of them clumped in the middle, with very few at the extremes.

If children are answering the survey, I would especially favor a smaller scale...

There are a variety of approaches you could use here, among others - it depends on which one, in the opinion of your dissertation advisor and yourself, is the most convincing...

(1) percentage of responses at 6 or above - is it significantly larger than 50%, which would suggest a higher proportion of "disagree" --> you can use the normal approx to the binomial here...

(2) is the mean significantly larger than 5.5 (the mid-point)? --> use a t-test

(3) is the median significantly larger than 5.5? --> use a Mann-Whitney test

Basically you may want to focus on showing that the data is mostly in the "disagree" zone of the scale.

Hope this helps.

JohnM
 
#5
I didn't want to use more than 5 scale rating but as discussed my tutor advised me that using 1 - 10 would be better!

Thank you for your advice I will go away now an look up your "testing" suggestions.

Thank you very much
 
#6
One last question, you have mentioned if the mean and median are higher than 5.5 to use the two different tests, what if they are lower than 5.5 can I use the same tests or will this mean I'll have to use something else.

Thank you
 

JohnM

TS Contributor
#7
It doesn't matter if the mean or median is above or below 5.5, you can use either test - you are actually just trying to test to see if the mean or median is significantly different from 5.5. Do the t-test for the mean, and the Mann-Whitney for the median.

The reason you would do both tests is that the t-test assumes that the data is pretty close to a normal distribution, and the Mann-Whitney doesn't require that assumption (doing analyses with two different methods is a good safe way to hedge your bets....)