Twenty-five items are a lot for a single scale, which is what you would get if you simply added all the items together. That is standard practice for creating a continuous variable from a series of ordinal Likert items, but it requires careful attention to whether thee items do indeed form a single scale. Assessing Cronbach's alpha is a first step in that direction.

Tools such as Mplus are allow you do much more complex analyses, and they are especially useful when you think that set of items measure two or more components of a concept (i.e., there are sub-scales within your measure).

But since it sound like you are new to the whole idea of creating scales, I would recommend starting with the simplest approach which is to assess the reliability of creating a scale through Cronbach's alpha. If you are using SPSS, you can find that under Analyses: Scales: Reliability.

One thing you may have to do before to use that command, however, is to make sure that you do not have any "reverse scored" variables. In other words, all the items have to have the same meaning for a higher score, so that they are all positively correlated. I would thus recommend that you examine the correlation matrix among the 25 variables to make sure that they are all coded in the same direction and thus positively correlated.

Once you have a set of positively correlated items, this site has a useful tutorial on Cronbach's alpha in SPSS:

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

Four options are

1 treat as a nominal category as you suggest - and use dummy variable coding you need to make sensible cutoffs and the number of categories can grow and become problematic if you are interacting this variable  with other variables

2 create an ordinal variable and use orthogonal polynomial coding that exploits this ordinality - this can be quite parsimonious

3 Treat as a numerical variable and fit say linear, quadratic and cubic model - looking for the most parsimonious fit - this model can be rather inflexible

4 Treat as a numerical variable and fit  underlying smooth relationship as in a General Additive Model which can be very flexible and parsimonious

Much will depend on the nature of your data and your experience.

In Spearman rank correlation, where one variable  is ordinal and the other, interval/ratio, you will convert the latter into ordinal.

In your example, when you categorize your data, you categorize both variables and you apply chi-square based measures of association such as Cramer's V or phi, or contingency. coefficient.

you have to use spearmans rank correlation for ordinal data

Standardize your Likert scale amd use Pearson corr.

You can group the score into categories and then assign each category the midpoint of its range as a score. The category for highest scores will still be "open-ended" but you can examine the distribution in that category to decide on the value for it.

Everyone, thanks for very helpful advises! Really appreciate your responsiveness! Following your recommendations I decided to categorise the responses having examined their distribution! Further I am going to use one way ANOVA which will allow me to see of there is any statistically significant differences among those experience categories.Running correlation analysis as I see is not appropriate. 

Normal practice is to treat likert scales as continuous variable even though they are not. As long as there are >=5 options, the bias from discreteness is not large.

I simulated the situation for you. I generated two variables with continuous random data from two normal distributions with a correlation of .50, N=1000. Then I created likert scales of varying levels from the second variable. Then I correlated all these variables with each other.

Correlations of continuous variable 1 with:

continuous2 0.5

likert10 0.482

likert7 0.472

likert5 0.469

likert4 0.432

likert3 0.442

likert2 0.395

So you see, introducing discreteness biases correlations towards zero, but not by much as long as likert is >=5 level. You can correct for the bias by multiplying by the correction factor if desired:

Correction factor:

continuous2 1

likert10 1.037

likert7 1.059

likert5 1.066

likert4 1.157

likert3 1.131

likert2 1.266

Psychologically, if your data does not make sense as an interval scale, i.e. if the difference between options 1-2 is not the same as between options 3-4, then you should use Spearman's correlation instead of Pearson's. However, it will rarely make much of a difference.

Here's the R code.

#load library

library(MASS)

#simulate dataset of 2 variables with correlation of .50, N=1000

simul.data = mvrnorm(1000, mu = c(0,0), Sigma = matrix(c(1,0.50,0.50,1), ncol = 2), empirical = TRUE)

simul.data = as.data.frame(simul.data);colnames(simul.data) = c("continuous1","continuous2")

#divide into bins of equal length

simul.data["likert10"] = as.numeric(cut(unlist(simul.data[2]),breaks=10))

simul.data["likert7"] = as.numeric(cut(unlist(simul.data[2]),breaks=7))

simul.data["likert5"] = as.numeric(cut(unlist(simul.data[2]),breaks=5))

simul.data["likert4"] = as.numeric(cut(unlist(simul.data[2]),breaks=4))

simul.data["likert3"] = as.numeric(cut(unlist(simul.data[2]),breaks=3))

simul.data["likert2"] = as.numeric(cut(unlist(simul.data[2]),breaks=2))

#correlations

round(cor(simul.data),3)

Can you Likert-scored variables go together to form a scale, or do you need to use them one at a time? If those variables are all supposed to be measures of the same thing, then you should test whether they can be combined. Are you familiar with Cronbach's alpha as a way to assess scales?

 I have created experience level groups as follow

1. 1-4 yrs

2.5-8 yrs

3.9-12 yrs

4.13-18 yrs

because I have exact indication of number of years it was easy to allocate the data into above groups for the consequent analysis on the means difference

Do you mean the assessment of the instrument's reliability. Yeah, I know and it seems to be fine :)

oh, do you mean that i should run some kind of test before grouping/allocating the way I did?  

It seems as if you want to transform the continuous variable, years of experience, into something ordinal, so that you can correlate it with other variables that are Likert scores, is that right?

If so, you are "losing information" in the years of experience variable by transforming it into a lower level of measurement.

The other alternative is to combine your Likert items into a single scale, which would be continuous, and thus gain you information. You can do this if the ordinal variables are all measures of the same theoretical concept.

In terms of reliability, you cannot assess the reliability of single variable, such as years of experience, unless you have additional data (e.g., a test-retest correlation). You can, however, use Cronbach's alpha to assess the reliability of a scale that is composed of multiple ordinal variables.

now I understand,  losing information.. as it is more precise ..

However, let me clarify:

Yes it is the same theoretical concept (25 items 7 point likert scale questionnaire used for its evaluation)

are you suggesting that I can use this as continuous data and correlate it with indicated years of experience? say pearsons correlation?

I am very sorry if I seem to ask too absurd qs :)

IF you have 25 items measuring the same underlying entity you could use factor analysis to get a better - (ie more reliably estimated) predictor variable- the package Mplus has very good facilities for this sort of latent regression analysis

THis book

Data Analysis with Mplus  http://www.guilford.com/books/Data-Analysis-with-Mplus/Christian-Geiser/9781462502455/contents

shows you how to do it and unlike the Mplus manual gives lots of help on interpretataion

Twenty-five items are a lot for a single scale, which is what you would get if you simply added all the items together. That is standard practice for creating a continuous variable from a series of ordinal Likert items, but it requires careful attention to whether thee items do indeed form a single scale. Assessing Cronbach's alpha is a first step in that direction.

Tools such as Mplus are allow you do much more complex analyses, and they are especially useful when you think that set of items measure two or more components of a concept (i.e., there are sub-scales within your measure).

But since it sound like you are new to the whole idea of creating scales, I would recommend starting with the simplest approach which is to assess the reliability of creating a scale through Cronbach's alpha. If you are using SPSS, you can find that under Analyses: Scales: Reliability.

One thing you may have to do before to use that command, however, is to make sure that you do not have any "reverse scored" variables. In other words, all the items have to have the same meaning for a higher score, so that they are all positively correlated. I would thus recommend that you examine the correlation matrix among the 25 variables to make sure that they are all coded in the same direction and thus positively correlated.

Once you have a set of positively correlated items, this site has a useful tutorial on Cronbach's alpha in SPSS:

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

If you are using SPSS you can read this book ''SPSS Survival Manual 4th edition by Pallant, J (2010)''. I once used it and it gives a good understanding of how you can collapse a continuous variable into groups or collapse the number of categories of a categorical variable and many more relevant topics so you can have good understanding and apply to your data .

When the 2 variables are continuous and met the assumptions of correlations (normality, linearity, no extreme values) we use Pearson correlation

When, the above assumptions violated, and or the variable is not continuous, then we can use Spearman rank-order-nonparametric test {which is typically used for ordinal with ordinal variables}

for likert scale, the items are ordinal, but usually we do summing for the items to get total score, which is considered as continuous variable. then we do Pearson correlation