Help deciding which statistical test to use in SPSS?

[helenoc24]

Hi All,

I am completing some research and wanted to double check that I am using the correct statistical test to analyse my data in SPSS.

I was to look at whether Voting in the community (Categorical response Y/N answer) impacts on the participants feelings that their views are being heard (Likert scale response SD/D/N/A/SA) AND their feelings of being part of the community (Likert scale response SD/D/N/A/SA). The data is normally distributed.

I was going to use a MANOVA but wanted to double check if this was correct before proceeding. Any help would be much appreciated and please let me know if I need to provide any further information.

Thanks,

Helen

 

[CB]

I was to look at whether Voting in the community (Categorical response Y/N answer) impacts on the participants feelings that their views are being heard (Likert scale response SD/D/N/A/SA) AND their feelings of being part of the community (Likert scale response SD/D/N/A/SA). The data is normally distributed.

Do you have just one Likert item for each construct, or Likert scales with multiple items?

The data is normally distributed.

It isn't, sad to say. Likert response scales produce a discrete distribution (e.g. if SD = 0, D = 1, N = 2, A = 3, and SA = 4, then only these 5 scores are possible). The normal distribution is a continuous probability distribution, so Likert scores logically cannot be normally distributed. However, this is not necessarily a problem for parametric analysis. E.g., as the sample size grows larger, the sampling distribution of the coefficients will approach normality, even if the conditional distribution of the response variable is not normal.

I was going to use a MANOVA but wanted to double check if this was correct before proceeding. Any help would be much appreciated and please let me know if I need to provide any further information.

You could, but I'm wondering: What valuable information is a MANOVA going to provide you with that you couldn't get from two separate, simpler, tests? (e.g., two independent samples t-tests, or two mann-whitney tests?)

 

[helenoc24]

Hi,

Thank you for you quick (and helpful) response.

I am a bit of a beginner when it comes to statistics and had assumed that the scaled data would be interval (which was one of the assumptions of normal distribution), reviewed the histograms and then the skewness/kurtosis z-scores which led me to the belief that the data was normally distributed. I can see what you have put below, thanks for your explanation. Do you feel that it would be best to go for a non-parametrics analysis or do the z-score and visual representation (histogram) mean that a parametric analysis could be completed?

I understand and agree with what you have said about using two simpler tests. To be honest, I was unsure of how many outcome variables I had and was struggling to decide on any test with certainty.

Thanks again,

Helen