Hello folks,

I am doing some aerospace research in the form of an online questionnaire on surveymonkey.com

The questions are mainly multiple-choice with some chose-the-three-most important-from-six ranking questions.

Although I understand the content of the survey, I need to do a critique of the significance of my survey and it results it produces such as number of respondents, discriminating between answers etc...

Does anyone have any experience of this or can they point me towards some resource or help online ?

If you are involved in the aviation or aerospace industry then please let me know as I would also appreciate your help as a respondent.

Many thanks
BigD

__________________
You make your own luck.

www.shootingthecheese.ch
The place for fun and chat in Switzerland.

I have an A-Level in Statistics, but with regards to Aerospace i know bugger all.

Not sure if i can be any help?

__________________
It is ridiculous claiming that video games influence children. If Pac-man affected kids born in the eighties, we should by now have a bunch of teenagers who run around in darkened rooms and eat pills while listening to monotonous electronic music....

Let me give you an example:

I have a survey question :

Choose three aspects of flying that you most enjoy:
1st choice 2nd Choice 3rd Choice

a) The throbbing engine
b) Arriving quickly
c) Less tiring
d) Cheaper
e) single mode of transport


How do I analyse this data ? How many respondents do need to make the mode significant ? What can I learn from the answers ? How should I treat invalid answers (say only two choices made) ?

This is the kind of thing I am trying to investigate , even before I get the results....

I am imagine this is all pretty standard marketing stuff...
BigD

__________________
You make your own luck.

www.shootingthecheese.ch
The place for fun and chat in Switzerland.

First of all you will need to pigeon hole as many different types of people as possible into different groups, then sub-divide these groups into smaller groups and then you can cross reference these groups.

Treat each question as a survey of its own and cross reference it to each group before you start finding trends between questions and of course an overall conclusion of the survey as a whole.

i.e

large groups:

everyone
men
women

smaller groups:

everyone single
everyone married
everyone young
everyone middle aged
everyone old
single men
married men
young men
middle aged men
old men
single young men
married young men
single middle aged men

etc...
etc...


As you can see these subgroups can go on forever, and we havent even started thing about where these people live, how much money they earn, etc... and all these groups need to be ratioed compared to how often they fly (or whatever the main premise of your survey is). This data will have to be obtained from a seconardy source.


For instance:

100 men where asked:
Q: Do you prefer orange juice or apple juice when flying?

A: Orange Juice - 65
A: Apple Juice - 35

100 women where asked:
Q: Do you prefer orange juice or apple juice when flying?

A: Orange Juice - 55
A: Apple Juice - 45


Most people would deduce from this that on average 60% of people prefer orange juice when flying. This would be a mistake, as im sure that the proportion of men flying and women flying is not 50/50. I have no idea what the ratio is but accordingly the answer for on average what percentage of people prefer orange juice when flying....it centainly wont be 60%.

Or another factor could be the fact that if you are surveying people on your contacts list in switzerland and england.........but you cant just add this data together as the swiss overall might fly more than the english.

The more and more one delves into the world of statistics the more factors and problems one encounters. And even if you spend 5 years working out everything possible and cross reference it all, the data is now out of date and therefore no longer relevant.


This is the reason why on the internet, newspapers, (and mainly the BBC bloody news) lots of crazy stats get mentioned, when in reality the data has been warped for either marketing purposes or just purely hype.


I did many projects at school showing how from one set of results on a survey can be analyzed to provide a vast array of completly different outcomes.


So really what you need to do is decide on your target specific group/demographic and just survey that group.


In terms of displaying the data, graphs and pie charts etc are a good way of seeing the big picture. this is when freak results at either end of the scale will scream out at you and can perhaps be taken out of the data

Hope that helps

cheers
SC II

p.s. I know what i have written above is mainly common sense, and im sure you are already aware of most of it so dont take it as me being condesending.

p.p.s Just ask as many people as you can, and however you analyze the data it wont be wildly wrong.

p.p.p.s. one thing is for sure that marketing people have no clue on even the basics of statistics let alone the complexities of estimation in sample design.

__________________
It is ridiculous claiming that video games influence children. If Pac-man affected kids born in the eighties, we should by now have a bunch of teenagers who run around in darkened rooms and eat pills while listening to monotonous electronic music....

Your questions are actually the more difficult ones on the trade. There is a lot of professionals that earn very nice sums of money to stet up surveys and its target groups. Unfortunately, I am not one of them.... Though I did a lot of surveys in a previous life, either we did a couple of a wild assumptions or handed over the job to the [market research] agency.

I am guessing your questionnaire will be sent to a "large" number of professional in your trade. Like, for example a customer data base? Or, are you sending it to the general public?

If you have are working with a customer database, then you have your sample. Depending on the number of answers, you can estimative the significance. This guys have a good website, and nice tools. Unfortunately, I was not able to find anything on line that had less superficial information. (we used to consider that 10% respondents out of 30k database were already good enough, with less than 5%, we would repeat it; this might be an empiric approach to the question)

Surveymonkey is a quite nice tool to group and analize segments of your audience, and discard I would recommend you add demographic questions - like gender, age, professional, or any other that might be relevant for your case. [Mind you, you will need to upgrade to the paying version to do this. Even converting to spreadsheet requires that you do an upgrade]

If you are sending it to the general public, how will you do that? Are you actually sitting on database, it will be something random (like picking names out of a phone book)?

I would also recommend you to do the same question in different ways, to ensure the result is not biased. Also, responders tend to lie (and will lie) and you should insert a number of questions to ensure that the are telling you consistently the same across the test.

For example, the one question you have given as example. I would maybe repeat those parameters as a form of a table: X axis with variants, Y with a rating scale. Rating should be a pair number of options like for example 1-very cold, 2-cold, 3-hot, 4 very hot. The reason for this is to force people into a decision - is it hot or cold? - and not to take refuge into a pleasantly warm choice. Also, when you are making comparisons, you should state against what you are comparing. Cheaper than...?

My 2 cents...

__________________
My blogs: about food and Zurich 1200 fountains

Can certainly verify that. I remember doing a telephone survey a few years back to which one question asked respondents what make of a certain tool they answered....about 30% gave a brand that didn't even make that particular tool.

Over the years I've totally lost what little faith I ever had in market research.

BigD, it's not the marketing people you need, it's market research or market insight people. They are the ones that do the analysis, marketing people just either use the results to define what's needed in the market, or bend the results to suit their campaign/launch/whatever.

__________________
Bring on Summer!

Many thanks for your input. It is actually very useful in a practical way. When I started out I was concerned that I wouldnt have enough to say about the data, but by cross-referencing "pilot think this, whereas airport managers think this", I have plenty of material.

The irony is that I have already spent five years getting this project off the ground

BigD

__________________
You make your own luck.

www.shootingthecheese.ch
The place for fun and chat in Switzerland.

Thanks Lucy for your input. May survey originally had about 36 questions with subsections. I was told my my prof that it needed to be no more than 20 question so I compromised on 25. Which means that there was not much scope for validation (asking same question in different way), although there is some overlap. The sample group are prequalifies as being within the industry. I currently have more than 50 respondents...

BigD

__________________
You make your own luck.

www.shootingthecheese.ch
The place for fun and chat in Switzerland.