Stats analysis help would be very much appreciated

#1
I have carried out some research. The research involved respondents filling in a questionnaire of 62 questions. Answers were on a scale - Strongly agree/agree/disagree/strongly disagree. My questionnaire was an attitude survey. ie respond to a set of statements such as "managment communicate regularly with staff about safety".

Now I need to analyse the results and I am just at a loss where to start with which statistcal tests etc. I have done a stats course but it was 14 months ago at the begining of my course and I found it difficult then. Please help! (I am using the SPSS database for analysis). I have created a results template today and entered all the results and I think I have done it correctly.

Are descriptives, frequencies and maybe multiple regression the way to go (please dont laugh too much if I am way out here).

I feel really stuck (total mental block) so any advice would be very much appreciated - thank you :)
 
Last edited:

bugman

Super Moderator
#3
Hi Oakmoss,

Im not very familiar with the analysis of survey data - this PDF should give you some ideas.

How many respondants were there to the questionaire?

Descriptives (plots and tables) would be a great place to start (frequency of responce to each caterogry for the most important questions). This might be all you need to do, depending on your objectives.

The scale data can probably be formatted into frequencies and analysed as such. However, hypothesis testing might not be the way to go unless you have specific questions you wish to answer.

Hope this helps in some way.
 
#4
So you made up some questions, have some data, and now what....?

been there, done that.

Like in the above response, do an explore in SPSS get lots of numbers table and stuff.

However, it's quite hard to make SPSS outputs nice. I use Excel. Put a table generated from spss, or generate it in excel and make a few graphs/charts with color and headings....etc...

now think back why you did this study. what did you want to find? Who actually wanted it in the first place? Are the people you describe this data to going to like what you say? Are they expecting a response, format...etc..?

You could do a few fancy things in spss, and when I mean fancy I mean not scientific but looks nice and is traditionally accepted by big bosses even though its not the right way to do it.

Just remember two things. Type and strength. What's the type of relationship you have in your data? How can you visualize or picture it, map it out. If you identify a type of relationship see where it leads to, perhaps a hypothesis test that bad atttiudes are related to such and such, vs. they are not related. Then pick a suitable test, make a decision rule if Pobs<.01 reject null and accept alternative, then estimate the strength....etc...

just take it easy and make the descriptive looks nice. you can always do a hypothesis test later on, and a lot of times showing the results are good enough. And when they ask what does it mean? Just say in the data we collected we found so and so, further studies are needed to determine the relationship and strength....

hope this helps,
jj
 
#5
Thank you so much both of you.

I had 158 respondents. It is a survey of attitude towards safety.

Thanks for pointing me in the right direction - much appreciated :)