SETUP

library( foreign )
library( memisc )
library( knitr )
library( dplyr )
library( xtable )
library( stargazer )
library( broom )
library( pander )

LOAD DATA

dat <- readRDS( "./Data/CompleteHazardSpells.rds" )

# lapply( dat, class )
#                   
# head( dat, 25 ) %>% pander

SPELLS

Inspect Spells

dat.sub <- dat[ ,c("ein","fisyr","age","FS_Totrev_adj","prof","JustNowProfessionalized") ]

head( dat.sub, 100 ) %>% pander

	ein	fisyr	age	FS_Totrev_adj	prof	JustNowProfessionalized
1	10211502	1999	1	77197	0	NA
2	10211502	2000	2	302158	1	1
10	10435507	2001	3	22254	0	NA
11	10435507	2002	4	38846	0	0
18	10482891	1999	1	48137	0	NA
19	10482891	2000	2	46809	0	0
20	10482891	2001	3	46344	0	0
21	10482891	2003	5	39729	0	NA
27	10494938	1998	0	0	0	NA
28	10494938	1999	1	0	0	0
29	10494938	2000	2	134188	1	1
30	10500759	2000	2	3964	0	NA
31	10500759	2001	3	2414	0	0
32	10500759	2002	4	2342	0	0
33	10500759	2003	5	5286	0	0
35	10503145	1998	0	30471	0	NA
36	10503145	2000	2	77109	0	NA
37	10503145	2001	3	230863	1	1
44	10506600	2000	2	0	0	NA
45	10506600	2001	3	38077	0	0
46	10506600	2002	4	31202	0	0
47	10506600	2003	5	29743	0	0
60	10510481	1999	1	0	0	NA
61	10510481	2000	2	0	0	0
62	10510481	2001	3	85853	0	0
63	10510481	2002	4	75063	0	0
64	10510481	2003	5	48988	0	0
65	10510909	1998	0	59352	0	NA
66	10510909	1999	1	51470	0	0
67	10510909	2000	2	56985	0	0
68	10510909	2001	3	55693	0	0
69	10510909	2002	4	62800	0	0
70	10510909	2003	5	62556	0	0
71	10511056	1998	0	21196	0	NA
72	10511056	1999	1	24823	0	0
73	10511056	2000	2	21049	0	0
74	10511056	2001	3	31252	0	0
75	10511056	2002	4	25123	0	0
82	10512664	2000	2	20980	0	NA
83	10512664	2002	4	20555	0	NA
84	10512986	2000	2	0	0	NA
85	10513088	1998	0	23433	0	NA
86	10513088	1999	1	21718	0	0
87	10513088	2000	2	17017	0	0
88	10513181	2000	2	8972	0	NA
89	10513181	2002	4	21996	0	NA
90	10513181	2003	5	69734	0	0
91	10513268	2000	2	33267	0	NA
96	10513496	1998	0	36494	0	NA
97	10513496	1999	1	44541	0	0
98	10513496	2000	2	46138	0	0
99	10513496	2001	3	45439	0	0
100	10513496	2002	4	50838	0	0
101	10513496	2003	5	56575	0	0
102	10513575	1999	1	1418	0	NA
103	10513575	2000	2	4418	0	0
104	10513575	2002	4	1220	0	NA
105	10513575	2003	5	1984	0	0
106	10513589	1998	0	52034	0	NA
107	10513589	1999	1	166141	1	1
111	10514230	1999	1	0	0	NA
112	10514230	2000	2	313509	1	1
113	10514747	1998	0	0	0	NA
114	10514747	2000	2	0	0	NA
119	10515050	1999	1	47215	0	NA
120	10515050	2000	2	43059	0	0
121	10515050	2001	3	95275	0	0
122	10515050	2002	4	58865	0	0
123	10515050	2003	5	34193	0	0
124	10515357	2000	2	3226	0	NA
125	10515357	2002	4	58926	0	NA
126	10515357	2003	5	44397	0	0
127	10515363	2000	2	62831	0	NA
128	10515363	2001	3	225484	1	1
146	10516071	1998	0	89105	0	NA
147	10516071	1999	1	224349	1	1
152	10516434	2000	2	11405	0	NA
153	10516434	2001	3	6983	0	0
154	10516434	2003	5	8940	0	NA
155	10516517	1999	1	31566	0	NA
156	10516517	2000	2	183391	1	1
157	10516579	1999	1	0	0	NA
158	10516579	2001	3	0	0	NA
159	10516646	1998	0	67462	0	NA
160	10516646	1999	1	63669	0	0
161	10516646	2000	2	194600	1	1
162	10517116	2001	3	1350	0	NA
163	10517116	2002	4	0	0	0
174	10517798	1999	1	38686	0	NA
175	10517798	2000	2	137693	1	1
182	10518008	2000	2	33966	0	NA
188	10518193	1999	1	0	0	NA
189	10518193	2000	2	132529	1	1
193	10518534	1999	1	0	0	NA
194	10518534	2000	2	209403	1	1
195	10518763	2001	3	6281	0	NA
196	10518763	2002	4	35310	0	0
197	10518763	2003	5	34798	0	0
198	10519140	1999	1	12807	0	NA
199	10519382	1998	0	55193	0	NA

Typical Spell Lengths

NOTE - do we want to experiment with imputing observations backwards for sensitivity analysis?

NOTE - if we don’t impute spell lengths backwards, should we set age to first year in dataset?

# count of spells by length

table( table( dat$ein ) )

## 
##    1    2    3    4    5    6 
## 1535 2384 1450 1010  857  508

# proportion that professionalize

sum( dat$prof ) / length(unique(dat$ein))

## [1] 0.3172779

Number that never professionalize - are right-censored because of end of study.

sum( dat$prof == 0 & dat$age == 5 )

## [1] 2815

# 2815 that never professioalize
# 7744 orgs

2815 / 7744

## [1] 0.3635072

Those that drop out before professionalization:

do.prof <- tapply( dat$prof, dat$ein, sum )

these.do.not <- names(do.prof)[ do.prof == 0 ]

dat.not <- dat[ dat$ein %in% these.do.not , ]

nrow( dat.not )

## [1] 15384

table( tapply( dat.not$age, dat.not$ein, max ) )

## 
##    0    1    2    3    4    5 
##  362  495  437  493  685 2815

2472 / 7744 # do not report (fail)

## [1] 0.3192149

2815 / 7744 # right-censored (alive in 2003 but small)

## [1] 0.3635072

2457 / 7744 # professionalize

## [1] 0.3172779

table( tapply( dat.not$age, dat.not$ein, max ) ) /

  table( dat$age )

## 
##         0         1         2         3         4         5 
## 0.1212324 0.1102695 0.1046456 0.1260870 0.2049671 0.9013769

Observation Accounting

table( dat$AGE )

## 
##    1    2    3    4    5    6 
## 2986 4489 4176 3910 3342 3123

table( dat$fisyr )

## 
## 1998 1999 2000 2001 2002 2003 
## 2986 4489 4176 3910 3342 3123

table( dat$age, dat$fisyr ) %>% kable

	1998	1999	2000	2001	2002	2003
0	2986	0	0	0	0	0
1	0	4489	0	0	0	0
2	0	0	4176	0	0	0
3	0	0	0	3910	0	0
4	0	0	0	0	3342	0
5	0	0	0	0	0	3123

table( dat$fisyr, dat$prof ) %>% kable

	0	1
1998	2986	0
1999	3873	616
2000	3494	682
2001	3411	499
2002	2990	352
2003	2815	308