Hi, Vaios,

The simplest way to do so is to recode the 7-points by merging the two lowest response categories and the two highest response categories. This will retain the ordinal ranking of responses; however, I suggest you reconsider this recoding. Reducing the variability of responses. Variability in responses is almost always a very nice thing to have in most statistical tests.

Peter

Hello Vaios,

I agree with Peter A Kindle that there's no practical benefit to collapsing a 7-point Likert-type response scale to a 5-point scale...especially since you've already administered the survey in that (7-point) form. At best, you'll have awkward scale values (e.g., if you fix 7 -> 5, and leave 1 -> 1, then 6 -> 4.29, 5 -> 3.57, etc.; each intermediate conversion is (x/7)*5, where x is the original value on the 7-point scale). In fairness, that's just a linear transformation, so no information is lost, nor do you change the shape of the distribution.

However, if you collapsed categories, you'd not only lose information but potentially change the shape of the distribution.

Good luck with your work!

Thanks all for your response, guess that you're both right that doing so gonna loose a lot of variability

If you did decide to go through with this (and as the above commentators say it is unclear why this would be "practical"), it is important not to treat the resulting values as if you administered a 5-point scale. If you just collapse 1-2 and 6-7 this supposes that these folks would have ticked the anchors on a 5 point scale.

It is of no benefit converting 7-likert scale to 5 likert scale. The respondents may have responded differently if it was originally 5 likert. It always good to consider data handling and analysis during the design of the study.

I think you have to keep it as a 7 point scale surely if you asked the participants using that scale.

Vaios G Kyriakis. Vaios, I think it might be helpful if you told us why you considered it might be good / necessary to reduce your scale from having 7 down to 5 response options.

Also, could you tell us whether you have already collected your data or not, please? If you have not yet collected your data, you face fewer problems.

Given the advice you have received up above (which I agree with), I am hesitant to tell you that it's easy to recode the data in SPSS because I think that your doing so might not be advisable. If you'd like me to tell you how to do it, however, please get back.

Vaios G Kyriakis , it is better to keep the 7-scale. Recode is simple, but you need to have strong reasons to do so. Let me guess what is the problem... you need to correlate that variable with other variables measured in a 5-points scale? Is this your case? If your problem is the diversity in scales, you can always think about to standardize your variables.

If you just want to convert the scale scored 1-7 into one scored 1-5, while retaining all the original information in the 7-point scale, you coud recode as follows:

1 --> 1.0, 2 --> 1.667, 3 --> 2.333, 4 --> 3.0, 5 --> 3.667, 6 --> 4.333, 7 --> 5.0. That gives the two scales the same range, retains near-equal intervals between categories, and does not discard any of the 7-category distinctions you asked respondents to make.

If you must (for reasons I and others find hard to imagine are persuasive) collapse the 7-category scale into just 5 categories, I don't care for combining the top two and the bottom two as was initially suggested. Instead:

1 --> 1, 2 --> 2, 3 -->2, 4 -->3, 5 -->4, 6 -->4, 7 --> 5.

That seems to me more defensible; it assumes respondents choosing the most extreme categories would do so regardless whether there are 5 or 7 to choose from, while respondents who want to answer medium-high or medium-low have two categories to choose from in the 7-point scale but only one possible category in the 5-point scale.

I dont see any reason of switching to a 5 point.In fact the reliability of a 7 point scales is more than 5vpoint.

Let's suppose that from a semantic point of view, option 2 is closer to option 1 than to option 3, and option 6 is closer to option 7 than to option 5, then recode:

1=1, 2=1, 3=2, 4=3, 5=4, 6=5, 7=5

If the case is that option 2 is closer to option 3 than to option 1, and option 6 is closer to option 5 than to option 7 (always from a semantic point of view), then recode:

1=1, 2=2, 3=2, 4=3, 5=4, 6=4, 7=5.

If "practical reasons" arise from low frequencies, take care of the meaning of the new categories.

Anyway, don't forget the above valuable comments.

I am going to conduct a survey. Some of the variables are of 5-point likert scale and some are of 7-point. I want to have a standard 5-point scale. Can I change the ones on 7-point Likert to 5-point? Will it affect the reliability?

Thank whole community for trying help with my question. I will consider all of your points before I made my modifications... All the Best.... Stay Safe and healthy!!!

Daniel Wright can you share the reference of the below mention statement, please?

If you did decide to go through with this (and as the above commentators say it is unclear why this would be "practical"), it is important not to treat the resulting values as if you administered a 5-point scale. If you just collapse 1-2 and 6-7 this supposes that these folks would have ticked the anchors on a 5 point scale.

If you want to compare results of different likert scale based questionnaires, you should normalize the data.

You should definitely not collapse the scale in a way that 1&2 as well as 6&7 mean the same as some people suggested. This is not what the participants intended when they decided to cross 2 instead of 1.

For the ones who cannot imagine, here is a "practical issue" for applying a 7 point scale and converting the scores into 5 points.

I will make a translation-adaptation study for a 7 item scale. I cannot decide on some translations for several items. As a result, I have a set of very similarly worded items.

I need to blend these items with the items of another 5 point scale, which is also going to be used for concurrent validity.

The 5 point scale has 10 items. Five of them are reverse coded.

It is traditionally reported over 100 points. The suggested method is adding 1/ subtracting 5 from item scores (regarding the coding direction) and multiply their sum by 2.5. It is the System Usability Scale by Brooke (1996), by the way.

On the other hand, I can use the same reverse/forward coding method by adding 1 / subtracting 7, and I use the multiplier 1.66666666667 to obtain a score over 100, without converting the scale.

A scale with more points is more "sensitive". In case that you are going to make any kind of factor analysis, it is not a good idea to convert a 7 point scale into 5 points.

A cool guideline by IBM for transformations.

https://www.ibm.com/support/pages/transforming-different-likert-scales-common-scale

7-point scales are a little better than 5-points—but not by much (Nunnally 1978). But, you should seek help from existing literature before forming your questionnaire.

If you are overwhelmed with low frequencies for some of the options, be practical. You can re-categorize the options depending on the closeness of the options (like what Jorge Ortiz Pinilla suggested here - https://www.researchgate.net/post/How_to_convert_or_deminish_a_7point_likert_scale_to_5point_likert_scale_in_SPSS).