Kaci,

We'd need a bit more information to help guide you. When you say you are doing a correlational survey does that mean you want to learn which responses are best correlated with one another? For starters, of course, you will want to determine Pearson's r values. You can also determine the statistical significance of these correlations. (Most software packages will automatically generate these with the Pearson's r results.) More generally, as you are probably learning, the analysis can get quite complicated. Have you done the study already or might there be time to tweak it in order to better answer the questions you are trying to answer? You mention there is only a small amount of data. Are you referring to the number of questions in the survey or the number of respondents? I could go on with questions about survey validation, etc. but that is enough for now. Feel free to post more details and with luck the group will be able to help guide  you a bit.

Gary

I have already done the study, and have entered the raw data into Excel. I bought the SPSS GradPak Software, but because I am still trying to figure out what to do with this data I am just using Excel for now. Basically what I am trying to do is this: I did a survey of Nurses' Self-Reported Self-Assessment of Biological Events. There is a questionnaire that consists of a 5 point Likert scale and then there are demographic questions: Education; specialty; years practicing as nurse; age; actual experience responding to disaster, etc. What I am attempting to do is determine if there is any correlation between the variables...eg....if education and actual experience responding to a disaster affects their overall self perceived self efficacy of biological preparedness. Does that make sense?

Kaci,  Small data sets sometimes require different statistical methods, in most cases, use of nonparametric statistical methods, like Chi Square comparisons across sub groupings on variables of interest.  If I were you I would begin with simple descriptive stats that characterize your universe of responses, ie., # respondents, and then medians and ranges on other variables, generally using the demographic variables as independent variables, which need to be subgrouped to find comparisons on each one with any of the dependent variables of interest.  For example, specialty could be subgrouped in ways that have a minimum of at least 5 cases in each subgroup, and then compare the subgroups on median splits constructed on any single dependent variable, where you find for any dependent variable a split point that groups the responses into an upper and lower half for that variable.  And to the extent the distributions are uneven showing a subgroup with possibly more responses in the upper half of the dependent variable you will be identifying nonrandom patterns about the influence of the subgrouping variable on the dependent variable, assuming if they were random, they'd all have similar 50/50 splits ...  To test this bivariate (two variables) distribution of response patterns, you could use a Chi Square test...  I would then proceed to do more bivariate comparisons and you may find some interesting things just doing these median splits across independent variable subgroupings, some of which could also be reconstructed with median or quartile or tercile splits depending on how the case counts per sub grouping work out...  Hope this gets you started...  You may find some examples of how to use quartile groupings and comparisons across independent variable sub groups here: 

Article Developing quality of acute care indicators through claims d...

Ok....I will have to take some time to digest all that, lol...remember...I have ZERO statistical background so what you just suggested was literally Greek to me. I think I get the gist of what you are suggesting; just going to have to play with it and see what I come up with. Thank you for the starting point....I have faith that I will eventually learn this language!

Hi Kaci,

I'm a director of statistical core at Nationwide Children's Hospital.  If you have funds we can do your statistics and write statistical part of your manuscript.

Thanks,

Igor

Hi Kaci, if you have some problems with medians and so on, you could start with a visualization of your data. You can plot a histogram and visualize how often two events appear similarly. Hereby you create a so called contingency table between two parameters. Next you may compare two parameters. You have ranked data and can compute the Spearman rank correlation. After this you can continue with a more sophisticated analysis.

Mathias is being helpful when he suggests using visualization - graphics of your data.  The kind of bivariate relationships I think are most natural to explore can be most easily seen by plotting the two variables of interest, in what are called scatterplots.  There are examples in the article I previously provided a link to... Usually the dependent variable (one of your likert scaled preparedness questions) is on the y axis, and the independent variable or it's sub grouped categories from your demographic domain are on the x axis.... look at examples and try to  build from them.  You can do that within excel...  Best of luck..  Just get started.  One thing will lead to another...

You guys are all giving me great info....Igor, I would LOVE to have someone do this for me, and as much as math and I do not get along, I would really like to figure this out so I am able to explain it and defend it...although, I may take you up on your offer if I can't make heads or tails of this. Even with the small amount of data that I have (I only had 75 responses, and some of them were incomplete)...looking at the raw data is overwhelming...not sure where to start. But I think with the suggestions that I've gotten, I may be able to at least begin to make some headway...I hope!

Ok Edwin, when referring to my Likert questions....I was going to use the Overall score as the one variable to compare with the other variables simply because it was a synopsis of the entire questionnaire...should I used each question, or would just using the summary of the entire section be sufficient? Basically it indicated that 47% of the nurses responding reported that they were somewhat unfamiliar with biological preparedness activities....

Hi Kaci,

You may want to grab a copy of:

Hulley SB, Cummings SR, Browner WS, Grady DG, Newman TB (2007): Designing clinical research. Wolters Kluwer Lippincott Willians & Wilkins

to form a well rounded perspective of your data and chose the most appropriate tool to answer your research question.

Thank you Herve...I will add that to my list, lol

Hi Kaci,

I am not recommending parametric regression analysis (which might be biased) because of the nature of your data: small size survey, where traditional statistics face certain challenges particularly in social, medical, and behavioral studies. This book by Paul R. Yarnold will help you:

http://www.amazon.com/Optimal-Data-Analysis-Guidebook-Software/dp/1557989818/ref=sr_1_3?ie=UTF8&qid=1417483087&sr=8-3&keywords=Paul+R.+Yarnold

It is quite easy to follow through, which comes with the software UniODA (univariate ODA). However, more advanced Multivariate ODA (automated CTA) has to be purchased separately. But in your case Kaci, I think UniODA will work just fine unless you have conducted surveys on multiple hospitals. Paul is also on RG with a lot of papers freely available to the public. Also check out his website with free journals on the ODA method: http://odajournal.com/

Best regards,

Richard

Kaci - please go and talk to someone. A few answers and pointers on ResearchGate are not a substitute for getting proper advice from someone who understands your data, your methods and your research question.

Flying blind is never a good plan, and calling out for help to passers-by is not a great option.

Ronan, I am talking to someone...I am asking the advice of people on Research Gate...their opinions are just as valid as anyone elses...thank you for yours

Excellent responses above and Ronan's advice should be heeded.  All I would add is:

(1) if data quantity is small it may not represent the population under analysis so you would need to caveat your observations accordingly

(2) when starting a new analysis I find it is usually helpful to chart the data graphically on time series or as a distribution of scores, as this can assist with initial qualitative analysis.

Good luck with it. 

Kaci,  You're on your way:  # responses = 75;  overall preparedness score = 47%  !!  That's work to build upon.  What's next?  Exploring components of the overall preparedness scores, where some may have been typically higher, and some were much lower??  I'd stick with the overall score description and further detail that:  What's the range of the overall scores?  clumped tightly around 40% to 60%? or spread out 0% to 100%?  And to be clear, Kaci, is the overall score a calculated score based on previous research with the questionnaire, or is it an average (sum of component scores / (divided by) # separate scores), or is it the median, the middle score dividing a respondent's scores into an upper and lower half?  What kind of picture, graphic image can you construct to show the distribution of the overall score - bar chart... line graph??  Work on that, further detailing the overall preparedness score, make sure you can show what your numbers mean to any audience with a picture of it, well labeled.  This is a key finding , answering what is likely a key question, like, "How prepared do nurses self report they are when considering biological emergencies"?  Detail this key finding, and then explore your other demographic variables to see how they might be associated with variation about the overall score.... You are analytically launched!  Keep it up ...

Hello Kaci,

As a recent graduate from Capella, I understand what you are going through.  I would suggest  to Google assistance with quantitative data.  It will recommend and provide you with information on resources of people who help with this.  When I finished my paper, I had to pay a lot of money to an editor who happened to know about data and got help from there also.  FYI you will need to get an editor besides the school editors.  I would also recommend a University or college close to you that may have students or teachers that help with reading quantitative data.  I am not an expert on quantitative data, but seek the resources, there are experts out their in which this is their field who are willing to help.  Be prepared it is costly?

Wilhemina....I do have a college here in town and was considering going to the Math department to see if anyone would be able to assist me.  I really wish I had taken it upon myself to take a Statistics class!! Thank you for the advise!

Kaci, please make sure that the techniques you use fit your research objectives and hypothesis testing, which is the most important no matter how complex or how easy the methods are.

Hi Kaci,

May be you can try GraphPad Prism to analysis your data, it is very simple and powerful for survival analysis.

http://www.hearne.com.au/Software/GraphPad-Prism/Editions#version_GraphPad Prism 6.0

Your question is way too general to be answered. We know only that we are in an observational setup, but know nothing of the field, the question(s) to be resolved, the survey design, the population(s), etc... So the only reasonable answer has been already given by someone who know what he's speaking about :

"I wish to perform brain surgery this afternoon at 4pm and don't know where to

start. My background is the history of great statistician sports legends but I

am willing to learn. I know there are courses and numerous books on brain

surgery but I don't have the time for those. Please direct me to the

appropriate HowTos, and be on standby for solving any problem I may encounter

while in the operating room. Some of you might ask for specifics of the case,

but that would require my following the posting guide and spending even more

time than I am already taking to write this note.

-- I. Ben Fooled (aka Frank Harrell)

R-help (April 1, 2005)"

Does your budget include paying for analysis?

Dear Kaci

You need to seek out a statistician (academic may be better in what I will suggest). ask them to undertake a limited analysis for you (there may be a charge) then explain the results to you.

Whether you have no statistical knowledge or very limited knowledge it would take at least a one semester course (for someone with a good educational background or level) to bring somebody to an appropriate level, to be able to tackle what you hint at. I would also be particularly concerned if the survey designers had no statistical knowledge (I assume that was not the case here?) 

Any approach you attempt, especially self-education, I think will be fruitless. Statistics is one of those disciplines which is well covered by the literature but you need a tutor to interpret the concepts and jargon, more importantly, for you, to guide on the process of analysis. In my experience the latter is not usually done well through reading/self-teaching.

A very good book for the lay-person  (who has a good educational background) is Discovering Statistics (using IBM SPSS Statistics) by Andy Field published by Sage (4th edition). Many readers view the approach in this text as humorous, easy to read and covering essential fundamentals whilst describing 'how to do' analyses. A glance at this will indicate the enormity of your task

I can help ..........

You simply need two columns of data then use a statistical programme that will do a correlation or regression. Or you can do it by hand (it's not that hard) or use an online statistics calculator for free :)