Cronbach alpha for Survey question with Select all that applies.

Question

Hi!

I have a survey questionnaire with 15 items, all of them are multiple choice questions. This is composed of 11 items with Select all that applies and 4 items with only to choose one question only. I was advised to do a cronbach alpha for it.

I normally know that typing the code for cronbach in SPSS, but How do I code the responses, especially with the questions with Select all that apply together with the questions with only to choose one answer only?

Can you also please give an example? That would help a lot!!

If you have any reference (books, links, etc) that may of help in my inquiry, that would be highly appreciated.

Also, if other reliability test is needed, kindly tell me. My goal is to test the reliability of the survey as a whole. Thank you!!

Answer 1

Hello,

Sorry for the late response. It is necessary to code the responses correctly when preparing your survey data for Cronbach's alpha analysis in SPSS, especially when dealing with the combination of "select all that apply" and "choose one answer only" questions. This is how the responses can be coded:

Coding Questions with "Select All That Apply"

Every option in "select all that apply" inquiries is handled as a distinct binary (0/1) variable. If a question has five possible answers, for instance, you will generate five different variables:

Question 1: Q1_1 (0 = not picked, 1 = selected) is Option 1.

Q1_2 is the second option (0 = not picked, 1 = selected)

Q1_3 is the third option (0 = not selected, 1 = selected).

Q1_4 is the fourth option (0 = not picked, 1 = selected)

Q1_5 is the fifth option (0 = not selected, 1 = selected).

Each question is coded as a single variable with categorical replies for "choose one answer only" questions. For instance, if there are four alternative responses to a question:

Question 12: Q12 (1 = Options 1, 2, 3, 4, 5 = Options 4)

As an illustration

Take into consideration a poll that looks like this:

Choose all three items that apply.

Select just one response (four alternatives).

Choose both of the applicable options.

Select just one response (three alternatives).

Coding Illustration:

Q1: (each classified as 0/1) Q1_1, Q1_2, Q1_3.

Q2: Q2 (possibly coded 1, 2, 3, or 4)

Q3: Q3_1, Q3_2 (with a 0/1) code for each

Q4: Q4 (either 1, 2, or 3).

Using SPSS to calculate Cronbach's Alpha:

Put Data in: Open SPSS and enter the coded data.

Analyze:

Go to Analyze > Scale > Reliability Analysis.

Transfer the variables to the "Items" box (Q1_1, Q1_2, Q1_3, Q2, Q3_1, Q3_2, Q4).

Choosing "Cronbach's Alpha" from the "Model" menu.

Press "OK" to initiate the analysis.

References

Book: Samuel B. Green and Neil J. Salkind, "Using SPSS for Windows and Macintosh: Analyzing and Understanding Data."

Laerd Statistics website a thorough manual on how to use Cronbach's alpha and other statistical studies with SPSS.

This should help you in accurately coding survey replies and running an SPSS Cronbach's alpha analysis.

I hope this helps. If you have any additional questions, please let me know. Take care,

Dr. Christal-Joy Turner

Answer 2

I am in the middle of statistics - so let me preface by saying this answer may not be an "Expert" answer - so if anybody has any corrections or noted I've made a mistake, please feel free to let me know.

Ok, so first we should note that the results of the Cronbach alpha will be a result between 0 and 1, with a higher alpha denoting an internal consistency among internal items. The Cronbach alpha is not the same alpha used in hypothesis testing as the threshold for rejecting the null.

There are common benchmarks for values, typically looking for .7+ for basic research, .8 for clinical observations. Now to your question.

For the coding, for multiple choice answers of a scale (rate satisfaction 1 to 5) will be coded to their answer - 1 point for the lowest unsatisfied value and 5 for extremely satisfied.

In the same survey, for the select all that apply, since every answer is as good as any other we can assign binary values to each - so if there are 5 options (4 choices and a select all), the first 4 will be worth 1 point, and the "select all that apply" would be worth 4. That would be the same if you picked all 4 of the first choices individually (1+1+1+1) or the "Select all" (+4).

So now for an example -

How satisfied were you staying at the Warewolf hotel?

Extremely spooktastic - 5 points

Good - 4 points

Mediocre - 3 Points

Poor - 2 Points

Terrible at terrifying - 1 Point

Which attractions did you find most exciting (Select all that apply)

Draculas Boatride: 1 Point

The Mummys Kitchen: 1 Point

In the Garage with Frankenstein: 1 Point

Pennywise Porch: 1 Point

Select All: 4 Points

Then you do the calculations - answers at 0 show no correlation, answers closer to 1 show high correlation. As mentioned before, the benchmarks are typically found at .7, so answers at this level or higher have a higher indications the measurements are reliable.

Hope this helps, this is as far as my understanding goes!