Hello Talkstats people,
I am another hopeless case who needs some help with my dissertation statistics. Basically I am not sure if the method outlined below is valid. What do you think?

For my dissertation I have surveyed student attitudes and taken diary recordings of behaviour with regards to water consumption. As such I need to test correlations between different category of data. There are 8 sets of data I will be comparing Those 6 sets from the survey - environmental attitudes, water attitudes, water knowledge, perceived externalities, barriers, and consumption behaviour. As well as those 2 sets from the diary for the control group and the group receiving and educational intervention.
Each category employs a mix of numerical, nominal and ordinal questions, each with a different total number of questions.

In order to transform the data to test correlations between categories I need to transform it. I am planning to assign a scores, from -1 to +1, to each type of answer given so each respondent will have a score for each category of data. Thus enabling me to use a Spearman Rank Correlation.

For example a question asking respondents to rank, among 10 listed issues that challenge society at present, those they deem the most important first. As such any who select an environmental issue for in their top 3 most important issues would be given a score of +1 and so on.

However, I am not certain if this is the correct way to transform the data and I believe it may be open to bias on my part. For example, regarding the question above, why only assign a score of +1 to people who ranked issues in the top 3 why not the top 5? Also will I have to justify how I have interpreted the results for each question in the dissertation write up, in order to make the method reproducible?

Also the total number of questions in each categories differs so the scores will differ. Is there any way I can negate this? Should I just give different weights to each data set when putting it into SPSS for the correlation tests?

Does my method of transforming the data make sense using a score system?


I don't have a strong maths/statistical background so this is all a bit of a mess, sorry if this seems a simple problem but it is a struggle for me.

Any help would be really appreciated.

Regards with peace and love on top,

Another Hopeless Case