# results ![[count-winsorize-0.95-clustse 5.png]] ![[frac-winsorize-0.95-clustse 4.png]] ![[sum-winsorize-0.95-clustse 6.png]] # probe interactions by dividing users into 5 bins (based on t0 variable) ## count badness Impossible to split into 5 **equally-sized** `t0` bins (what we preregistered) because most users shared 0 or very few bad tweets (most users would end up in bin01), so I created 5 bins with **equal range** instead. - y-axis tick labels: bin number (no. of users in the bin) (t0 range] ![[dv_pc1_badness_threshold40_count_winsorize0.95_impute-zero_se-clustblock-expose-expose_adj_count.png]] ![[dv_pc1_badness_threshold65_count_winsorize0.95_impute-zero_se-clustblock-expose-expose_adj_count.png]] ### pilot campaign binned results ![[dv_PC1_badness_threshold40_count_winsorize0.95_impute-zero_se-clustblock-expose-expose_adj_count 1.png]] ![[dv_PC1_badness_threshold65_count_winsorize0.95_impute-zero_se-clustblock-expose-expose_adj_count 1.png]]