Quantitative, inductive survey analysis?

[Burnsie_UK]

I will soon be analysing my survey (I posted another thread about this, but I feel I over complicated it, so deleted it).

Essentially, my survey has >100 responses. It is inductive. I don’t know what I’m going to find, I am not testing a hypothesis.

the survey questions can be broken down into two simple groups; demographics (people) and the views of said people.

an outline of the questions and the data type are below.

I have generally only done deductive statistics previously, using experimental methods. Im therefore a little lost on inductive research.

I'm looking to try and find any patterns from the below. I.e. if qualification level is linked to some of the options that people put first in one of the ranking questions. Whether years’ experience means they are more or less likely to have heard of the product / know someone that has. Whether someone that thinks the biggest drawback is time (question 5), also says they think they need the most time to do the thing (question 6).

It could be that none of the inductive examples above would actually pull out in statistics….

however, to start with I am just looking for views about how to analyse?

Any help or guidance would be great


Demographics
1. Age – free number (nominal – continuous)
2. Sex – check box choice (nominal)
3. Qualification level - check box choice (categorical)
4. Years’ experience - check box choice (categorical)
5. Current role - check box choice (categorical)


Views
1. Have you heard about X? – Yes/No choice (categorical)
2. Have you ever used X? Likert (ordinal – once coded)
3. Know anyone else that uses X? Likert (ordinal – once coded)
4. Overall benefits of X? Likert (ordinal – once coded)
5. Overall drawbacks of X? Likert (ordinal – once coded)
6. How much time you should have to complete X? - check box choice (categorical)

7. Detailed benefits of X – Rank 1-7 (ordinal)
8. Detailed drawbacks of X – Rank 1-7 (ordinal)
9. Detailed facilitators of X – Rank 1-7 (ordinal)

10. Confidence of using X – Likert (ordinal – once coded)
11. How to improve confidence– Rank 1-4 (ordinal)

 

[Karabiner]

What is the title of your study, and what are the research questions?
What will the results be used for, which problems can be solved with them?

With kind regards

Karabiner

 

[Burnsie_UK]

What is the title of your study, and what are the research questions?
What will the results be used for, which problems can be solved with them?

With kind regards

Karabiner

Hi, thanks for the swift reply.

the overall project is looking to development of a novel field-based lower-limb neuromuscular screening tool for use by technical coaches within female grass-roots youth soccer.

The survey is the first study.... before I make something new, I want to know what is already known/thought about the topic by the end-users themsevles.

Therefore the 2 research questions are

1. What are the potential barriers and facilitators regarding the implementation of a neuromuscular screening tool for use by grass-roots soccer coaches?
2. What is the current knowledge, perceptions and confidence levels of grass-roots soccer coaches regarding the utilisation of a neuromuscular screening tool?

If I know what the barries etc are, then we can try and develop a tool that takes these into account etc.

hope this helps

 

[Karabiner]

You could first describe first each variable as is (e.g. for age: minimum and maxium, median,
standard deviation; Have you heard about X: %yes and % no; detailed benefits: mean or
median etc.). You could also use graphics for this (e.g. Box-and-whisker plots for "age"; bar charts
for the frequencies of answers to items like "5. Overall drawbacks of X? Likert"

Then you could describe bivariate associations between Demographics and the Views.
For example, are those who heard about it younger than those who didn't. You can use graphics
(like box-and-whisker plots), and descriptive statistics (mean, median, standard deviation etc.).
Or, which median (or mean) repsonse to " Have you ever used..." was seen for different roles
or different qualification levels.

An important decision here is whether the Likert-type response scales should be treated
as ordinal, or may as interval scaled.

With kind regards

Karabner

 

[Burnsie_UK]

You could first describe first each variable as is (e.g. for age: minimum and maxium, median,
standard deviation; Have you heard about X: %yes and % no; detailed benefits: mean or
median etc.). You could also use graphics for this (e.g. Box-and-whisker plots for "age"; bar charts
for the frequencies of answers to items like "5. Overall drawbacks of X? Likert"

Then you could describe bivariate associations between Demographics and the Views.
For example, are those who heard about it younger than those who didn't. You can use graphics
(like box-and-whisker plots), and descriptive statistics (mean, median, standard deviation etc.).
Or, which median (or mean) repsonse to " Have you ever used..." was seen for different roles
or different qualification levels.

An important decision here is whether the Likert-type response scales should be treated
as ordinal, or may as interval scaled.

With kind regards

Karabner

Hi Karabner

Once again, thanks for taking the time to reply. That all makes sense..... however.... could you provide anymore detail / signposting on the last comment re how to treat the Likert reponses? are you getting at parametric vs non-parametric?

 

[Karabiner]

Some would argue that 7 point rating scales are ordinal, and therefore parameters
like the mean or coefficients like Pearson's correlation cannot be calculated. Some
others would argue that you can treat them like an interval scale, if the levels can be
assumed as evenly spaced.

In the present study, maybe either treatment wouldn't do much harm.

Why did you mention parametric or non-parametric - those terms refer to statistical
tests of inference. Do you consider doing such tests?

With kind regards

Karabiner

 

[Burnsie_UK]

My Likert have 5 points

The rankings vary depending on the number of barriers etc listed in some research papers.

I mentioned para / non para due to getting confused about the ordinal vs interval comment.

I have seen people use chi-square, anova, t etc to look at surveys like mine, but they, I am guessing, are testing a hypo, whereas I am not!?!

 

[fed2]

you could use the pearson correlation, invented by the great eugenicist, and sometimes statistician karl pearson.

 

[Karabiner]

I have seen people use chi-square, anova, t etc to look at surveys like mine, but they, I am guessing, are testing a hypo, whereas I am not!?!

Well, I don't know. From your description, I thought you just do a survey and
describe the results, but I may be wrong. With 5 point scales, and with ordinal
variables such as "Detailed benefits of X", non-paramatric tests would be obvious.

With kind regards

Karabiner