Bookmark and Share

Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

st: Piecewise regression using panel data


From   Sunkee Lee <[email protected]>
To   statalist <[email protected]>
Subject   st: Piecewise regression using panel data
Date   Mon, 17 Mar 2014 02:32:03 +0800

Hello. I need some help on doing piecewise regression using panel
data. Could someone please give me some help?

Currently, I am working on a project that investigates the effect of
salary system change on sales employees' performance. In particular, I
have a data set which has the performance data of sales employees
selling merchandises online. The business model of the company is
identical to those websites in the US that sells flash deals online.
Hence, the merchandises that the sales employees sell in my data are
products such as food, electronics, clothes etc. that are sold at a
very high discount rate for a fixed amount of relatively short time.
Originally this company's salary system was one that gave monetary
incentives to the employees proportional to their sales performance.
Therefore the better deals they source (and the more sales they make),
they would be paid more accordingly. In this system the fixed baseline
pay is very low. However, in the middle of my dataset, the company
changes the salary system into a fixed salary system in which the
baseline pay is much higher but there are no incentives related to
performance. I am examining how this change might affect the learning
rate of the sales employees. A basic assumption is that as experience
increases, sales employees learn from their experiences. Hence,
performance should increase according to experience level. Under the
incentive system, employees will be eager to learn from their
experiences and therefore the learning curve will be positively steep.
However, after the pay change, they would be less motivated to learn
and therefore the slope of the learning curve will decrease
substantially. My regressions results using a fixed-effect GLS shows
that this is actually true, and the effects are moderated by whether
the employee was a high performer or low performer before the change.

Now what I want to do is to graphically depict this change in slopes
and intercept using STATA, but I am very confused how this could be
operated. Below are a few of my problems:

1) I would like to do a piecewise regression. However, can the
-mkspline- command be used for panel data? Or a more general question
would be can piecewise regression be done using panel data?

2) If piecewise regression can be done using panel data, I wonder how
I could adjust for the intercept at the knot (changing point). I tried
using the -hascon-option, and it doesn't work. I assume it is because
it is panel data..

3) The relation I am looking at is the relation between accumulated
experience (number of prior deals that employees sold online) and
daily revenue per employee. Accumulated experience should be the
x-axis and daily revenue should be the y-axis. However I noticed that
in order to do a piecewise regression, the x-axis should be time
instead of experience since I want to look at the slope before and
after the change. Is there any way to reconcile this issue? Employees
join the company at different points of time so it is more confusing.

Thank you in advance for helping me out!
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/


© Copyright 1996–2018 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index