[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

From |
Ngoc Anh Vo Thi <[email protected]> |

To |
Statalist <[email protected]> |

Subject |
st: Problem of convergence with maximum likelihood estimation |

Date |
Mon, 10 Mar 2008 02:02:38 -0700 (PDT) |

Dear all, I'm re-posting a message posted some days ago. I�m conducting a study on the comparative cost efficiency of two types of foreign banks - those that enter a domestic market by setting up a greenfield investment (GRs) and those that enter by mergers and acquisitions of domestic banks (M&As), and want to find out which factors explain discrepancies in their efficiency. To this end, I�ve tried models suggested by Hung-Jen Wang in his paper �Heteroscedasticity and Non-Monotonic Efficiency Effects of a Stochastic Frontier Model� published in the Journal of Productivity Analysis (2002) by running his Stata program as provided in his website. That means I tried to include factors that are likely to affect the inefficiency term (u) by parameterizing mu or usigma or both given the general form of the distribution of u: N+(mu, usigma). However, the problem is these models barely converge with my data. I would like therefore to aks you why this is the case? Is it meant these models are not compatible with my data? Or is it because my zlist is too long (6 variables)? I would really appreciate your helps. Here is what I type for the estimation. *define variables in global macros, so they can be referred to easily later global yvar lncostn global xlist Y1 Y2 P1 P2 Y1Q Y2Q Y1Y2 P1Q P2Q P1P2 Y1P1 Y1P2 Y2P1 Y2P2 lnequit country_2 country_3 YEAR_2-YEAR_11 /* translog cost function with exogenous factors influencing the shape of the production function */ global zlist lnassets shareinv shareoff sharedep merger mergage **USE THE FOLLOWING FOR M&As VS GREENFIELD **use the procedure suggested by Wang (2002) *Run an OLS on the frontier-only function to get consistent estimates *(except the constant). They are used later as initial values. capture reg $yvar $xlist if foreign==1 mat b0 = e(b) /* record the coefficient vectors in a matrix */ *defining the maximization problem sfmodel $yvar if foreign==1, cost dist(truncated) frontier($xlist) mu($zlist) usigmas($zlist) vsigmas() ** (optional) supply initial values sf_init, frontier(b0) mu(0 0 0 0 0 0 0) usigmas(0 0 0 0 0 0 0) vsigmas(0) ** (optional) Search for better starting values for the specified variables. ** Based on Stata's -ml plot- command. sf_srch, n(2) frontier($xlist) mu($zlist) usigmas($zlist) fast /* "n(2)" cycles through the varaible list twice.*/ /* "fast" may speed up the drawing of graphs */ *do the maximization, with the gradients printed ml max, diff gtol(0.001) gradient *calculate inefficiency index, and possibly the marginal effects sf_predict, bc(bc_Wang) jlms(jlms_Wang) marginal /* Battese & Coelli (1988) efficiency index in bc_Wang; */ /* Jondrow et al. (1982) inefficiency index in jlms_Wang; */ /* "marginal" calculates the marginal effects of the zlist varaibles */ Thanks, Anh ____________________________________________________________________________________ Never miss a thing. Make Yahoo your home page. http://www.yahoo.com/r/hs * * For searches and help try: * http://www.stata.com/support/faqs/res/findit.html * http://www.stata.com/support/statalist/faq * http://www.ats.ucla.edu/stat/stata/

- Prev by Date:
**st: R: trend analysis for survey data** - Next by Date:
**RE: st: modifying a macro containing variable names (problem solved)** - Previous by thread:
**st: trend analysis for survey data** - Next by thread:
**st: Instrumental Variables with constraints.** - Index(es):

© Copyright 1996–2024 StataCorp LLC | Terms of use | Privacy | Contact us | What's new | Site index |