[R] Difficult survival model

[Daniel Malter]

Hi all,

I would appreciate your advice how to model the survival model for the
following data, especially if it can be modeled in one model or if I should
(have to) break it up (i.e. assume that some events are independent of each
other, etc.).  Data is on an experimental stock market simulation. Dependent
variable is the hazard of transition from either holding the stock to
closing the position (-1) or from not holding the stock to opening a
position in that stock (1) in the variable CHNG. Main variables of interest
are the regressors IV1 and IV2 (which are covariates, not factors). BUY and
SELL indicate the number of stocks of type STOCK that were purchased/sold in
ROUND. Accordingly, NBT and NET record the number of STOCK held at the
beginning and the end of period ROUND, respectively. HP is a 0/1 dummy
indicating whether or not shares of STOCK were held by an investor in any
given period.

And these are the issues I see in the data:

1. Over time, the stock may be held several times. This suggests a
conditional hazard. TIMESHELD measures the number of times a stock has been
held (meaning that the stock position must have been closed inbetween).
Alternatively, it could measure the number of prior transactions in the
stock (all kinds). This I intend to model with strata.

2. I also have repeated measures for individuals. As some investor may be
more of a trader than others, I want to model this with a random effect
(frailty).

3. Different stocks may have a different risk of being traded. This I would
model with another frailty term (or would I rather model it differently?).

4. In case that an investor holds a stock, there are competing risks between
increasing or decreasing the position.

5. There are also competing risks, so to speak, between the different
stocks. That is, when an investor makes a buy or sell decision, s/he decides
between all available stocks. One stocks competes with all other stocks to
be purchased. When selling a stock though, one stock only competes with the
other different stocks in the portfolio to be sold, of course.

I have no idea how to model 4. and 5. though and if this is possible. I
would also appreciate your feedback on the suggested way of modeling 1., 2.,
and 3.

Apologies for the lengthy description and thanks much for your support,
Daniel


ID	ROUND	STOCK	BUY	SELL	NBT	NET	IV1	IV2	HP
CHNG	TIMESHELD
1	1	A	0	0	0	0	3	4	0
0	0
1	2	A	10	0	0	10	3	5	1
1	1
1	3	A	0	0	10	10	5	4	1
0	1
1	4	A	0	10	10	0	2	1	1
-1	1
1	5	A	0	0	0	0	3	4	0
0	1
1	1	A	20	0	0	20	2	5	1
1	2
1	2	A	0	0	20	20	4	5	1
0	2
1	3	A	0	0	20	20	2	3	1
0	2
1	4	A	0	0	20	20	5	3	1
0	2
1	5	A	0	20	20	0	4	5	1
-1	2
2	1	B	15	0	0	15	1	2	1
1	1
2	2	B	0	0	15	15	1	2	1
0	1
2	3	B	0	0	15	15	2	2	1
0	1
2	4	B	0	15	15	0	5	1	1
-1	1
2	5	B	0	0	0	0	4	4	0
0	1
2	1	B	5	0	0	5	3	4	1
1	2
2	2	B	0	0	5	5	5	2	1
0	2
2	3	B	0	5	5	0	4	2	1
-1	2
2	4	B	0	0	0	0	4	1	0
0	2
2	5	B	0	0	0	0	4	1	0
0	2


PhD Program Strategy
Dept. of Management and Organization
Robert H. Smith School of Business       
University of Maryland
Van Munching Hall                                           
College Park, MD  20742
www.rhsmith.umd.edu
www.umd.edu

mailto:dmalter at rhsmith.umd.edu
mailto:daniel at umd.edu