|
|  |
[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]
Re: st: Re: questions about xtprobit
If I am understanding you correctly Nicola, the time variable you
propose would only capture a linear trend.
To give you an example of when such a linear trend term would not be
as appropriate, consider partisan dominance of U.S. state
legislatures. National shocks that result from a national political
scandal or an unpopular war could influence the popularity of local
candidates from the political party tainted by such shocks. A linear
time trend variable or even the addition the square of such a trend
would not capture such idiosyncratic shocks, but year-specific
dummies would. More generally, two-way fixed-effects designs are
useful because they eliminate higher level but otherwise unmeasured
effects that may influence all cases.
Dave Jacobs
At 11:34 AM 1/1/2007, you wrote:
I think you can use dummies for years. But why not using -xtlogit,
fe- with a time variable indicating the calendar year?
Nicola
At 02.33 30/12/2006 -0500, you wrote:
>ate: Fri, 29 Dec 2006 12:41:09 -0500
>From: David Jacobs <[email protected]>
>Subject: Re: st: questions about xtprobit
>
>What if one wants to control for time in a fixed-effects logit
>analysis? Suppose I have pooled time series with panels by year and
>I want to estimate using a two-way fixed-effects design. I
>understand that it is incorrect to use dummies for cases, but is it
>incorrect to enter separate dummies for each year in fixed-effects
>logit models?
>
>Dave Jacobs
>
>At 06:11 AM 12/29/2006, you wrote:
>>Not sure if -xtprobit- is the best choice, (I'm not an expert, but
>>it sounds that the choice depends on what "after a certain event"
>>means, i.e. are all observations subject to the same event at the
same time?).
>>Turning to technical questions on Stata, to my best knowledge you
>>cannot have fixed effects for -xtprobit-. But you can use
>>CONDITIONAL fixed effects for -xtlogit-. Remember that you cannot
>>use dummies to estimate logistic fixed effects (Hsiao 1986). You can
>>have exponentiated coefficients (odds ratio indicating the
>>percentage change in the probability of the outcome (the dependent
>>variable = 1) given a one unit change in the independent variable)
>>with option -or-.
>>Nicola
>>
>>At 02.33 24/12/2006 -0500, you wrote:
>> >Hi all,
>> >
>> >For my nmaster degree thesis, I have data on firms borrowing from many
>> >sources as shown below. I am trying to run probit model to see how
>> >likely it is that firm will have better access to finance (to
>> >different sources e.g. banks, bond market) after a certain event. With
>> >the data structure below, I want to control for firm's characteristic
>> >too.
>> >
>> >firmid financesource improveaccesstoloan x1
>> >1 1 1 8
>> >1 2 0 8
>> >1 3 1 8
>> >
>> >where x1 = indepedent variable capturing firm characteristics for each
>> >firm.
>> >I am thinking about using xtprobit for the following model:
>> >
>> >prob of improving access = financesourcedummies + x1+ x2+.....
>> >
>> >Is xtprobit the right command to use?
>> >Can I use fixed effects with xtprobit?
>> >I tried it and getting coefficient greater than 1. how do I obtain the
>> >correct marginal effects on probability?
*
* 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/
*
* 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/
