I am trying to establish if there is an association between two variables. My Outcome Variable, sum_anx_7 , indicates the level of Anxiety. My Input variable is the number of traumas experience in childhood, ACE_FREQ_SUM_agg.
summary(dt$sum_anx_7)
Min. 1st Qu. Median Mean 3rd Qu. Max.
7.00 8.00 14.00 16.54 24.00 49.00
summary(as.factor(dt$ACE_FREQ_SUM_agg))
0 1 2 3
561 278 88 91
According to the test Overview bellow, i should be opting for Spearman Rank in order to assess if there is a relationship.The result is significant.
Spearman's rank correlation rho
data: dt$ACE_FREQ_SUM_agg and dt$sum_anx_7
S = 157966345, p-value = 0.001171
alternative hypothesis: true rho is not equal to 0
sample estimates:
rho
0.1015935
However, I would like to have some more details. I was thinking of switching the outcome and input variable. So instead of sum_anx_7 ~ ACE_FREQ_SUM_agg, i would be looking at ACE_FREQ_SUM_agg ~ sum_anx_7. That way i could be using the Kruskal Walis test and potentially running ad-hoc pairwise comparisons. I am aware that anxiety cannot be causing traumas, but these type of test only asses the association right. Is it acceptable, from the statistical point of view to do this?
