The Likert scale?

[Tom James]

Hi there,

My name is Tom and I am really new to Stata and I am looking for a brianbox to help me figure something out! I am sure this question will be most mundane for many of you out there! I am looking to input questions such as "do you trust in medical research?" into STATA and then have options "strongly agree" all the way to "strongly disagree"? As much as I try and figure it out I think some of us just are not cut out for it! Please could someone offer me an insight?

Many thanks!

[wbuchanan]

Tom James it isn't very clear what you are wanting to do. Do you already have these data? If so, you can import them using one of several different commands for importing data. If you are trying to set up a survey to collect responses, you'd probably want to look at a different tool. If you are trying to input a set of data manually, there would need to be more information (e.g., an infinite number of points lay between strongly disagree and strongly agree so you would need to provide information about the response set).

[Nick Cox]

Billy (William => w) is right. But here is a guess.

You have 5 point scales. Enter your data as integer codes 1 to 5 and then define and attach value labels.

This is well documented e.g.

help label

Thousands of people have learned Stata. It seems now that people try to learn Stata by Googling quasi-randomly. That is not, in my view, the best way to learn Stata.

http://www.stata.com/manuals14/u.pdf

contains a very good grounding.

Start at the beginning, skim and skip what is obvious or appears irrelevant, and stop when it gets too difficult.

Keep going back to it as you gather more experience. What was difficult will start seeming obvious and what was very difficult will start seeming interesting and useful. It will still be true that bits can be skimmed and skipped.

Throughout, learn help or search in Stata before trying the internet.

[ericmelse]

Dear Tom,

I suppose you should also read the Stata Manual chapters on Item Response Theory and work through the examples presented there, simply type in the Stata command window:

help irt

as well as do some background reading on the subject.
One recent presentation from a Stata meet up you can find here.

[Tom James]

Dear all,

Thank you for your responses, and time!

Sorry I didn't convey it very well, I really had no idea, I have a book which I'm going through called "getting started with STATA" it's a tad old though, so I will definitely check out the manual!

Nick Cox you guessed right, I have a patient survey, which has this question which says "Please select the answer which best describes how you feel about each statement" this is followed by 20 statements each with strongly agree, agree, neutral, disagree, strongly disagree. I just am struggling to figure out how to set up a variable for which, when I get my patient data I can automatically just type in the responses as 1-5? I will have 200 patients you see and I wasn't really sure the best way I could go about doing it!?

thank you ericmelse & wbuchanan also for your help!

[wbuchanan]

Tom James if you're able to avoid hand entry that is always best and Stata isn't the best tool for data entry (although it is certainly possible). You could set up a do file like:

. clear

. input str3 id int(i1 i2 i3 i4 i5 i20)

            id        i1        i2        i3        i4        i5       i20
  1. "  1" 1 4 2 3 5 5
  2. "  2" 1 3 3 2 4 1 
  3. end

. 
. la def responsesets 1 "Strongly Disagree" 2 "Disagree" 3 "Neutral" 4 "Agree" 5 "Strongly Agree"

. qui: ds, not(type string)

. foreach v in `r(varlist)' {
  2.         la val `v' responsesets
  3. }

. li

     +----------------------------------------------------------------------------------------------+
     |  id                  i1        i2         i3         i4               i5                 i20 |
     |----------------------------------------------------------------------------------------------|
  1. |   1   Strongly Disagree     Agree   Disagree    Neutral   Strongly Agree      Strongly Agree |
  2. |   2   Strongly Disagree   Neutral    Neutral   Disagree            Agree   Strongly Disagree |
     +----------------------------------------------------------------------------------------------+

This assumes there are no other numeric variables in the dataset, but you could just as easily loop over the relevant variables using a forvalues loop if the item number appears in a consistent location in the variable name.

[Clyde Schechter]

One slight tweak to Bill Buchanan's code. -label values- accepts a varlist, so the loop isn't needed:

quietly ds, not(type string)
label values `r(varlist)' responsesets

And from the pedantry corner, I wouldn't use "Neutral" as the label for response 3 on a 5-point Likert scale. A person may well choose that response, not because they feel neutral about the matter, but because they are ambivalent--they may agree with some aspects of the proposition and disagree with others. Or they may be unsure whether they agree or disagree on balance overall. So I was taught that the midpoint of this scale should be "Neither Agree nor Disagree." Since that label tends to be too long to display nicely in most Stata output, I often shorten it to just "Neither."

[wbuchanan]

Clyde Schechter I hadn't looked at the help file for label in so long that I was completely unaware of that. Thanks for pointing it out.