diff --git a/LearnJulia/src/amtrak.jl b/LearnJulia/src/amtrak.jl
index 0269c1e..6071de6 100644
--- a/LearnJulia/src/amtrak.jl
+++ b/LearnJulia/src/amtrak.jl
@@ -4,6 +4,7 @@ using Cascadia
 using DataFrames
 using DataFramesMeta
 using Dates
+using Statistics
 
 url = "https://juckins.net/amtrak_status/archive/html/history.php?train_num=97&station=&date_start=07%2F01%2F2024&date_end=07%2F31%2F2024&df1=1&df2=1&df3=1&df4=1&df5=1&df6=1&df7=1&sort=schDp&sort_dir=DESC&co=gt&limit_mins=&dfon=1"
 
@@ -31,11 +32,19 @@ end
 
 
 mod_df = @chain df begin
-    @rsubset :act_dp != ""
+    @rsubset :act_dp != "" && :s_disrupt != "SD"
+    @select Not(:comments, :s_disrupt, :cancellations)
     @rtransform _ begin
         :act_dp = Time(:act_dp, dateformat"HH:MMp")
         :orgin_date = Date(replace(:orgin_date,  r" \(.*\)" => ""), dateformat"mm/dd/YYYY")
         :sch_dp = DateTime(replace(:sch_dp,  r" \(.*\)" => ""), dateformat"mm/dd/YYYY HH:MM p")
     end
+    # @rtransform  :delay = canonicalize(Dates.CompoundPeriod(:act_dp - Time(:sch_dp)))
+    # @rtransform  :delay = canonicalize(:act_dp - Time(:sch_dp))
+    @rtransform  :delay = :act_dp - Time(:sch_dp)
 end
 
+
+Statistics.mean(mod_df.delay)
+mod_df.delay
+Statistics.mean(1:20)