Likert

You should think in terms of measuring independent (predictor) variables and dependent (outcome) variables. For both the independent and dependent variables, you need to make arguments for their reliability and validity. The best way to achieve reliability is to use multi-item scales, would would be composed of several Likert-scored items, rather than just one.

Quantitative is to analysis data which collected using questionnaire

I concur with David. You need to determine which variables you're going to use. Also, if you're looking at relationships, then use some type of regression.

I definitely concur with David. I always learn something from his comments. Adding up further, I could see following two main variables in your study:

1. The innovative character of visualization/infographic

2. the perception of the professionals about the reader

Further you have divided professionals into three groups to examine the difference in perception.

Apparently you have decided that first is DV and the latter is IV. Thereafter, the very first thing is to decide about how variables will be measured and this decision will affect on your decision to select an appropriate analysis method (statistical method).

As per your statement you are thinking to make use of Likert scale type items to measure perception of professionals (IV). As David said, use multiple items to measure the perception. So you IV will be a continuous variable.

How do you plan to measure innovative character of visualization/infographic? Again, if you are going to use Likert type scale, use multiple items. Once you decide about how you will measure you DV, consult following link to decide about type of test you will need: https://stats.idre.ucla.edu/other/mult-pkg/whatstat/

For the 2nd hypothesis, you can simply run ANOVA to see if there is difference in perception of three groups of professionals. I would highly suggest to run a post hoc test to identify where exactly the difference, if any, is coming from.

If you have both IV and DV on a continuous scale, you can run the model in SmartPLS SEM as well. Though SEM is recommended for complex models, having several constructs, I don't think there is harm in examining interaction between two variables using SEM. If you opt SEM, SmartPLS's latest version is not available for free, however, you can make use of v2.x, which is freely available from their site.

I hope my reply was helpful.

Cheers!

Likert sacle belongs to ordinal scale. In two ways you can express ur results. One way is to present which one is most common against one statement like strongly agree / disagree (how much percentage). Another way is to analyse the scores by median and IQR, followed by application of non-parametric tests. If scores are found to be normal, then u can also go for parametric tests.

Ali, Indranil and David, thank you very much for the comments.

Ali Farooq:

about how to measure the innovative in visualization, I thought about the selection of media that have this feature (for example, those media that stand out in awards such as SND's Best of Digital Design, Malofiej, Data Journalism Awards). Thinking about it, I can assume that The New York Times, The Washington Post and The Guardian are more innovative than the Folha de S. Paulo, from my country, by the amount of awards. I can plot by the amount of prizes. In this case, I would consult professionals of these various medias and compare according to the position of the awards. Is it possible?

Hi Rodrigo.

The influence of perception on innovative characteristics can be measured by conducting regression, where perception is the independent variable and innovative is the dependent variable.

And I think you can initially compare the differences in perceptions between professional by conducting ANOVA/t-test (dependeing on number of groups). The results of the differences are the basis for regression.

You can refer to existing scales or develop your own measurement to measure perception. However, seriously consider the reliability of the scales, as you will have respondent from the UK and your own country cultural differences can result with serious reliability issue. Pilot testing is highly recommended.

Concerning your innovative measure, I think it is possible. Although there will be argument whether the award is expert- or public-based. But I don't think it is a major issue.

I hope this will give you some idea.

A lot of different statistical methods can produce the same results. (Personally, I would probably start with crosstabs to get a feel for the data, and then move to regression analysis.) The more important issue is defining the key variables, developing appropriate measurements, and constructing the predictive model. These steps are likely to determine whether your research succeeds or fails.

You should try to develop a complete model, not looking just at innovation and perceptions of the reader. A complete model will have greater predictive power, increasing its credibility, and will lessen the chance that the relationship you find is really due to a correlation with some third (and missing) variable. For example, that might include the designers' levels of skills, the tools they have available, and whether or not they work in a team.

Of course, the nature of the variables you create will affect what statistical approaches are available. For example, if your measure of innovativeness is dichotomous (yes/no), then you would need a tool such as logistic regression. Ideally, you would create a interval-level measure, but that may not be practical (or feasible).

A questionnaire is simply your SOURCE of data and it should not preclude you from using any statistical test. That decision is based upon such factors as scaling (e.g., ordinal or nominal), sample size, and design (e.g., separate groups vs. repeated measures). You can use a questionnaire as part of an experiment (just randomize the distribution of two different versions of the questionnaire). The one thing I do not recommend using with a questionnaire is open ended questions in hopes of doing a textual analysis. Do a synchronous interview (preferably face to face) if you want to do that sort of research. In order to find quantitatively scorable response formats, I suggest you look at major polling organizations (e.g., Gallup, Barna, Zogby, Mitofsky, NORC) and use their time tested phrasings.

I think you need to tell us more about how you measured the innovative characteristics and the perceptions before we can help you determine which were the independent and dependent variables.

Your data has been quantified. So you can rank your variable based on relative ranking and conduct rank correlation

Which method of data analysis should I use for a study using questionnaires?

Based on your depiction in the question - you can conduct 2 types of data analyses depending on the data you'd collected & your hypotheses formed:

1) Regression (can be multiple regression or even structural equation modeling if the relationships are more complex)

Reason being saw you have some variables that you intend to measure their correlationship e.g.

"My hypothesis is that the innovative character of visualization/infographic is directly related to the perception of the professionals about the reader (when they think their readers are more able to understand the grammar of the visualization, they are those that bet more on innovative visualizations)"

2) Analysis of Variance (ANOVA)

Reason being you might want to compare groups about their perceptions toward something e.g.

"other hypothesis: that different professionals (journalists, designers and programmers) have different perceptions. That is, I must work with three groups of professionals."

Pertaining to the use of Likert scale - yes some researchers / scholars treat it as ordinal scale (that require the use of non-parametric statistics). However some researchers also treat it as interval scale that justify to use parametric statistics. You can see RG link below on the debate:

https://www.researchgate.net/post/I_have_a_variable_that_is_measured_through_5_point_likert_scale_Should_it_be_considered_as_a_continuous_or_categorical_variable

Just suggestion - if you want to ensure your Likert scale data being accepted as internal scale data for parametric statistics tests, you can adjust your questionnaire to change from Likert scale to Semantic Differential scale using Bipolar Adjectives - you can refer to my post below how to change that:

https://www.researchgate.net/post/What_should_I_used_to_analyse_a_likert_scale_a_parametric_test_or_a_non_parametric_test

It all depends on how you want to analyze the data, whether it is to show what numbers represent in mathematical or statistical terms or how dependent and independent variables are related, the method is quantitative. But if the analysis will depend on the opinion of the researcher, the method is qualitative

Likert

Quantitative analysis is suggested for data collected through questionnaires.

I agree very much with Dr. Ali Farooq.

Use can get very useful details from the book given below. This is freely available in the internet.

Sekaran, U., 2003. Research methods for business: A skill building approach. John Wiley & Sons.

I just throw in one point: when doing statistical analyses about a question of "how and why" as it is the case in your "perception" task, it is dangerous, because you just want to measure the relatedness. In our case, it would be helpful to see how the people perceive and what they feel is important. Therefore I recommend that you additionally conduct some interviews to get a real insight in how they see it. Questionnaires can guide and influence people as they have just the choice to tick boxes (Likert skale), but they can not choose their own preferences freely - so you miss what might emerge from your sample.

Hope my additional view helps you.

Dear Rodrigo Cunha ,

First fix or set up the objective(s) of study specifically and clearly.

Then collect the relevant data.

Only then the job of selection of data will arise.

I agree with Dhritikesh - still we have qualitative aspects - perception - in your topic, so consider that when shaping your research objective and research questions.

Thanks to Sandra Schlick for recommending my answer.

Appreciation to the answer given by Sandra Schlick.

Thanks to Rodrigo Cunha for recommending my answer. Best wishes to him.

I agree with the view of Sandra Schlick as mentioned in her answer.

Thanks to her for acknowledging me.

The data analysis will depend on the type of questions in the questionnaires and whether they are quantitative or qualitative in nature. The right data analysis will depend on a careful review of the questions, where you will select either quantitative data analysis or qualitative data analysis. Both data analysis can require the use of analytic software. For instance, quantitative data analysis will require analytic software, such as Excel, SAS, SPSS, R, Minitab, etc. Correspondingly, qualitative data analysis will require analytic software, such as NVivo, ATLAS.ti, etc.

Type of data snalysis will depend on your objective of the study. Is it correlationsl, causal etc

I concur with David’s response but I would like to explore further. The type of analysis would depend on the nature of the questions in the questionnaire. Are the questions open-ended or closed ended? If the questions are close-ended then the Likert Scale approach can be used and quantitative (descriptive and inferential) analysis can be applied. However, if the questions are open-ended then either quantitative or qualitative analysis can be performed. In using the Likert Scale, the scale can be converted into scores to aid quantitative inferential analysis such as correlation and regression. Kothari (2004) provides detailed explanation on how to convert the scale into scores. For instance if you use a five point Likert scale and your questionnaire has 10 question or variables you can convert this into scores in a way that each respondent is given a single score that ranges between 10 and 50. The least score will be 10 (1x10) and the highest score will be 50 (5x10). Thus, if a respondent’s answer was 1 in each of the 10 questions, the total score becomes 10 and if another respondent’s answer was 5 in each of the 10 questions, the total score becomes 50. SPSS Transform Tool can be used to convert scale into score for quantitative analysis.

Philip Danquah is correct.

I agree with him.

Thanks to him for his reasonable answer.

Peter Asare-Nuamah has given a explanatory type answer. I also agree with this answer. Thanks to him for his explanatory answer.

The opinion given by Zainudin Awang is correct.

I already added this comment.

Thanks to him.

The suggestion of analyzing the responses using the graded response model or the graded unfolding model given by Alvin Vista is, as I understand, is a good and fruitful suggestion.

yes i definitely concur Ali, there is no harm, you can test the hypothesis using SEM.

Thanks to Sandra Schlick for recommending my answer.

Best wishes to him.

Thanks to Manabhanjan Sahu for his response.

Please be more specific when seeking advice on methodology. The choice of methodology depends on what specifically you want to do - test a simple hypothesis or conduct a comprehensive analysis.

Which method to use would depend on what hypothesis you are testing not what scale you used to measure your variable.

I am also of the similar view as mentioned by Raj Chawla. Thanks to Raj Chawla.

Which method to use would depend on your objective of the study As well as what hypothesis you are testing.

A. S. M. M. Hoque is correct.

I agree with him.

Thanks to him.

@Anuradha, thank you so much for the reference.

Regression analysis is an option. This type of analysis can be done.

Thanks to Anuradha Iddagoda for recommending my answer.

Best wishes to her.

The choice of statistical test is not dependent on the researcher's choice rather on data type, variable type, the question to be answered, and level of depth the researcher want to get. The size of sample size and number of questions asked in questionnaire is also very critical to select appropriate test.

Mubashar Majeed Qadri is correct. I agree with his opinion.

My view is that statistical analysis of data is superior to the grey data analysis.

If you use structure questionnaire, then it might be appropriate for you to use SPSS, AMOS, PLS, or some other relevant quantitative tools for analysis.

It depends on your hypotheses or research questions.

I do agree with Philmore Alleyne .

@ Mohd Ikhwan Aziz ,

Each of SPSS, AMOS and PLS, is a software for numerical computations only. These are not , in true sense, tools for analysis of data.

It depends on the type of variables -especially dependent variables! If you have continuous dependent variable, OLS regression is recommneded. For categorical dependant variable use logistic regression and for count variables Poison regression is suggested.

There are many other conditions involved in it too. However, the answer to your question would be it’s not the questionnaire that would decide type of analysis technique but the variables! Hope it answers your question.

Both statistical and grey data analysis can be used for this .

Major concern: your sampling sounds very heterogenous (journalists, designers, programers) that is worse than apples & oranges, more like apples, fish sticks, and donuts. If you have a large enough sample size, compare these three with ANOVA or Kruskal-Wallis. If you throw in "education" you have too many ways of measuring that variable: level, major, institution. Unless we are talking n > 500, I am worried about small group sizes.

If this is for a dissertation and you want something do-able and quick, here's what I would rather see: pick one profession (e.g., designers). Pick a measurable predictor or two (e.g., the trait of openness, years of experience in the field). Sort the data by years of experience. Run a Pearson r between openness and Likert. Do this in JASP and select Bayesian Sequential Analysis.

H1: The higher the level of openness, the higher the Likert rating.

H2: The strength of this correlation diminishes with more years of experience.