Here are some helpful tools:

https://statistics.laerd.com/spss-tutorials/cronbachs-alpha-using-spss-statistics.php

https://scholarworks.iupui.edu/bitstream/handle/1805/344/Gliem%20&%20Gliem.pdf?s..

Alpha only calculates the reliability of a scale, and most questionnaires are not scales. A scale is a set of items that measure the same construct and uses the same response categories throughout. (Alpha does not measure 'scaleness', for which you need to calculate a measure like Loevinger's H.)

So unless your questionnaire fits the above description, alpha is of no use to you. 

So while the tools mentioned above are useful, they are useless if they are wrongly applied. 

Ronan's remark on scaleness is correct and one should not use alpha on Likert-type items, it can be used as an approximation. In a syntax you could use:

RELIABILITY

/VARIABLES=vr1 vr2 vr3 vr4 vr5 vr6 vr7 vr8 vr9

/SCALE('scale 1') ALL

/MODEL=ALPHA

/STATISTICS=DESCRIPTIVE SCALE ANOVA TUKEY

/SUMMARY=TOTAL.

I have added ANOVA TUKEY as TUKEY will test if the items can be summed to one score. If the test shows no departure from additivity one can use ALPHA to calculated reliability. It does not ensure either scalability or unidimensionality. Factoranalysis can be used to explore unidimensionality. Ask SPSS for a one factor model and see how it fits (use ML as estimation procedure). Another option would be to use IRT models like Rasch or Mokken or .. These are not included in SPSS, can be found in R or in the integration of R with SPSS (run R from within SPSS). If you add R to SPSS Rasch models are added to your Menu.

Thanks Guys

I disagree with Ronan and Peter, since alpha is actually the most commonly used measure to assess the reliability of scales, and Likert-scored items are the most commonly used measures for alpha.

Because there have been so many questions about this topic, I have gathered a set of resources on it:

https://www.researchgate.net/post/How_can_one_help_people_create_simple_scales_from_Likert-scored_items?_tpcectx=qa_overview_asked&_trid=lxx80F2Qeehlh9Y1xKoutA62_

'Commonly used' does not mean that it is correct. More and more researchers tend to use IRT models. Because they can deal adequately with ordinal measurement level used in research. I use all kinds of methods to have good scales. I do not search for good scales, I create them or refine them. One of the things I always investigate for a scale is differential item functioning (DIF). I need to be sure that DIF is not present if I compare groups. If DIF is present for sex, it is not a difference in magnitude, but in quality. This type of result can not tested or found 'just' using alpha.

More and more does not many. My suggestion to Basile would be to look at the literature in his area and see what methods his peers are using.

And if DIF is an essential issue, then I would recommend Structural Equation Modeling.

If a researcher is using SPSS, following Youtube shows how to calculate the Cronbach Reliability:

https://www.youtube.com/results?search_query=SPSS+cronbach+calculation

Albeit Cronbach‟s Alpha is widely used as an estimator for reliability tests, it has been criticized for its lower bound value which underestimates the true reliability (Peterson, R.A. and Y. Kim, 2013). Composite Reliability can be used as an alternative as its composite reliability value is slightly higher than Cronbach‟s Alpha whereby the difference is relatively inconsequential (Peterson, R.A. and Y. Kim, 2013).

Following 2 links show how to calculate Composite Reliability (CR) in SmartPLS & SPSS AMOS (note: for the 1st link SmartPLS CR see slide 9)

https://www.google.com/?gws_rd=ssl#q=3-1+-+How+to+use+SmartPLS+software_Assessing+Measurement+Models_3-5-14-1

https://www.youtube.com/watch?v=zJQswqqTtNQ

This site may be help you

https://www.easycalculation.com/statistics/reliability-coefficient.php

I think we can all agree that Cronbach's alpha is a reliability tool utilized to determine if a set of items that measure the same construct (scale) is reliable. If your goal is to test the reliability of items that factor together well then Cronbach's alpha reliability tool will be useful. After completing a factor analysis/principle component using SPSS you will:

Click on:

analyze

scale

reliability analysis

alpha (default)

then click ok to determine reliability

using SPSS, Click on:

Analyze → scale → reliability analysis → alpha (default) → then select the items from left and send to right → click statistics → under descriptive, select scale and scale if deleted → click continue → then click ok to determine reliability.

Remember, Cronbach's alpha value should by ≥ 0.70.

Thanks.

 First, please realize that alpha is not  a measure of reliability, but rather of internal consistency.  Even then, it does not show that the test measures one construct, for it is quite easy to have a high alpha although the items reflect 2 or more independent factors.  For a discussion of the problems with alpha, please see my Psychometrika paper with Rick Zinbarg (Revelle, W.  and Zinbarg, R.E. 2009, Coefficients alpha, beta, omega and the glb: comments on Sijtsma.  Psychometrika, 74, (1)  p 145-154.  DOI 10.1007/s11336-008-9102-z).

Then, if I have not dissuaded you from wanting find alpha, computationally it is very easy to do in R.  Say you have your fifty questionnaires in a tab delimited Excel file.  Copy them to the clipboard and then into R:

library(psych)    #make the psych package active 

my.data

It's true, Crohnbach's alpha is only a measure of internal consistency. Reliability is best measured using a form of factor analysis. The type you select is dependent on whether or not the items you are using in your survey are from an existing index or not.

Cronbach's alpha may not be a direct measure of reliability, but it does set a lower bound on reliability. That is why you get suggestions that alpha should be above a certain cut-off level, such as .7 or .8.

In other words, if you get a high level of alpha, then you can be assured that your scale is indeed reliable.

If you want a point-estimate for reliability one could use Mokken-scaling. Although it is a technique for ordinal items I found it scales in general are a subset of items found using factor-analysis or with using alpha or a reliability analysis in SPSS. Mokken can also provide you with a scalability index. You can also use it to have a scale that has a pre-specified level of scalability. It does item trace line to be monotonically increasing, but does not demand tracelines to have a specific form (as Rasch-type model demand).