Write down an analysis plan/ desсrіption for how you did the analyses.
If this involved selecting your final model based on excluding variables until ass in the model were significant – change this to instead compare overall models to make your choice (F-statistic and adj R-square values).
See if there is an easy way to render your outcome normal (log, square root, square, exponential). If not then go as is but where you are left with just one predictor in your final model run a non-parametric model on that to give reassurance that the association is robust to breaking normality assumptions.
You must come up with the scatter plots to show the patterns graphically
Avoid ANOVA and run simple models (backward elimination/forward linear regression)- and describe the steps in a textual format .
Log transform outcome variable and run Pearson′s correlation
The only significant observation (15-19 e cig in 2014 vs 20-29 infertility in 2019) may raise a couple of issues:
The data on e-cigs in 2014 may not be as objective compared to e-cig data in more recent years
The biological plausibility/prevalence of male infertility in such a young age of 20-29 may not be totally justifiable!
Anyway, follow the above steps and ensure that the methodological approach is appropriate, interpretable and biologically plausible.
the above are the comments given by the supervisor for my dissertation for the analysis. i have attached a corrected work but not sure if i was able to follow all instructions. the analysis should be done in spss and and should cater all the comments of the supervisor. i would need analysis as per the instructions of the supervisor