TCAD Data EDA

Summary of the Data

• How many missing (value = -99 or blank entry): missing_# (counts)
• How many unique values: unique_#
• Note LAD can take values like “Unknown” or “Unknown but loaded”, although no missing is reported in the table below.
• Note CST = 2 for all incidents in the data

The summary table, named “Table_summ_0429.csv”, can be downloaded from google drive.

View Missing Patterns

The following variables have more than 7000 records missing.

The plot below shows the missing pattern among the above variables, where the 16943 rows indicating records in our database, columns indicating variables, and cells are colored according to missingness: black = missing and grey = present. The plot, named “Rplot_miss_pattern.png,” can be downloaded from google drive.

Understand missing STC

How missing is coded in LAD and STC?

                     
                       -99  -98   15 4906610
  Chlorine               1    0    0       0
  Corn Syrup             2    0    0       0
  Not Applicable         0    3    0       0
  Potassium Hydroxide    1    0    0       0
  Unknown             8433    1    0       1
  Unknown but Loaded     0    0  657       0

• Three obs with STC tag “-99” but known LAD tags: “Chlorine” (obs 7295) and “Corn Syrup” (obs 10641, 10642).

• Two obs with LAD tag “Unknown” but known STC tags. Likely the two incidents should have LAD tags “Ethylene Oxide” and “Sulfuric Acid”, respectively.

The R output shows the number of incidents with missing STC along years.

1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 
  71  117   94  121  166  274  259  359  501  408  521  434  383  355  304  270 
1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 
 296  323  282  220  267  265  226  244  271  263  203  242  352  227  214  193 
2012 2013 2014 2015 2016 2017 2018 
  69   91   64   45   33   36   36 

There are 8437 records with STC missing (coded as -99), most of which have INC starting with “AAR” or “DTE”. In the whole database, there are

• 6242 records with INC starting with “AAR” (among which, 4425 with STC missing),

• 2454 records with INC starting with “DTE” (among which, 2333 with STC missing).

INC entry that begins “DTE” (for Date To be Established) basically means that the accident’s DOA is unknown.

Mismatch between DOA and DOA.1

There are five records whose DOA and DOA.1 are different.

Mismatch between DOE and DOE.1

There are 1820 records whose DOE and DOE.1 are different (including the 1st record, “5557”, in the table above). Among them, most have DOE latter than DOE.1

• Three (which actually correspond to the same incident) have DOE 256-day earlier than DOE.1;

• One has DOE 182-day earlier than DOE.1;

• One has DOE 5-day earlier than DOE.1;

• 494 have the two dates 1-day apart;

• 177 have DOE 2-day latter than DOE.1;

• 258 have DOE 3-day latter than DOE.1;

• 886 have DOE more than 3-day latter than DOE.1;

These 1820 records, ordered by the difference between the two dates, are in file “subset_data_DOE_mismatch.csv” in google drive.