Hi There,
I was wondering if anyone could possibly help me with a couple of questions?
>
> I'm running some simulations in a do file, producing a binary string.
> The model itself is not binomial, and I'm looking at how well the tests
> pick this up.
>
> I'm running some correlation stats on the binary string (of wins and
> losses). The analysis including:
> reg w l.w
> reg w l.l.w
> reg w l.w l.l.w
> reg w l.w l.l.w l.l.l.w
> runtest w, threshold(0) , where w=1 is a win, l=0 is a loss.
>
> I'm looking to see whether the stats pick up that
> P(w_t|w_{t-1}) > P(w_t|l_{t-1}) significantly or not.
>
> The stuff i get out is for example
>
> . *** The Analysis ***
> .
> . reg w l.w
>
> Source | SS df MS Number of obs
> = 99
> -------------+------------------------------ F( 1, 97)
> = 1.21
> Model | .305025768 1 .305025768 Prob > F
> = 0.2740
> Residual | 24.442449 97 .25198401 R-squared
> = 0.0123
> -------------+------------------------------ Adj R-squared
> = 0.0021
> Total | 24.7474747 98 .252525253 Root MSE
> = .50198
>
> ------------------------------------------------------------------------------
> w | Coef. Std. Err. t P>|t| [95% Conf.
> Interval]
> -------------+----------------------------------------------------------------
> w |
> L1. | .1110204 .1009069 1.10 0.274 -.0892519
> .3112928
> _cons | .44 .0709907 6.20 0.000 .2991031
> .5808969
> ------------------------------------------------------------------------------
>
> . reg w l.l.w
>
> Source | SS df MS Number of obs
> = 98
> -------------+------------------------------ F( 1, 96)
> = 0.64
> Model | .163265306 1 .163265306 Prob > F
> = 0.4241
> Residual | 24.3265306 96 .253401361 R-squared
> = 0.0067
> -------------+------------------------------ Adj R-squared
> = -0.0037
> Total | 24.4897959 97 .252472123 Root MSE
> = .50339
>
> ------------------------------------------------------------------------------
> w | Coef. Std. Err. t P>|t| [95% Conf.
> Interval]
> -------------+----------------------------------------------------------------
> w |
> L2. | .0816327 .1017001 0.80 0.424 -.1202405
> .2835058
> _cons | .4489796 .0719128 6.24 0.000 .3062337
> .5917254
> ------------------------------------------------------------------------------
>
> . reg w l.w l.l.w
>
> Source | SS df MS Number of obs
> = 98
> -------------+------------------------------ F( 2, 95)
> = 0.73
> Model | .37218136 2 .18609068 Prob > F
> = 0.4832
> Residual | 24.1176146 95 .253869627 R-squared
> = 0.0152
> -------------+------------------------------ Adj R-squared
> = -0.0055
> Total | 24.4897959 97 .252472123 Root MSE
> = .50385
>
> ------------------------------------------------------------------------------
> w | Coef. Std. Err. t P>|t| [95% Conf.
> Interval]
> -------------+----------------------------------------------------------------
> w |
> L1. | .0930626 .1025876 0.91 0.367 -.1105994
> .2967246
> L2. | .0702372 .1025662 0.68 0.495 -.1333824
> .2738569
> _cons | .4090956 .0843447 4.85 0.000 .2416502
> .576541
> ------------------------------------------------------------------------------
>
> . reg w l.w l.l.w l.l.l.w
>
> Source | SS df MS Number of obs
> = 97
> -------------+------------------------------ F( 3, 93)
> = 0.91
> Model | .688662088 3 .229554029 Prob > F
> = 0.4413
> Residual | 23.5587606 93 .253320006 R-squared
> = 0.0284
> -------------+------------------------------ Adj R-squared
> = -0.0029
> Total | 24.2474227 96 .25257732 Root MSE
> = .50331
>
> ------------------------------------------------------------------------------
> w | Coef. Std. Err. t P>|t| [95% Conf.
> Interval]
> -------------+----------------------------------------------------------------
> w |
> L1. | .0961765 .1032586 0.93 0.354 -.1088746
> .3012277
> L2. | .0708528 .1034346 0.69 0.495 -.1345478
> .2762533
> L3. | .096522 .1032192 0.94 0.352 -.1084509
> .3014948
> _cons | .3644247 .0945632 3.85 0.000 .1766409
> .5522085
> ------------------------------------------------------------------------------
>
> runtest w, threshold(0)
> N(w <= 0) = 50
> N(w > 0) = 50
> obs = 100
> N(runs) = 45
> z = -1.21
> Prob>|z| = .23
>
>
> 1) I'm believing L1 to be the first order serial correlation statistic
> - (the probability of a win(1) after a win(1) this period) minus (the
> probability of a win(1) after a loss(0) last period)...i.e if L1=0 the
> terms are independant. Am i right?
>
> 2) To work out the second order serial correlation statistic do I add
> the L1 and L2 terms together? I'm looking for the prob of a win(1)
> after 2 previous losees(1's) compared to a win(1) after two previous
> losses(2's).
>
> 3) Why is the L2 from reg w l.l.w different from the L2 in reg w l.w l.l.w?
>
> 4) Is the P>|t| the p-value?
>
>
> 5) Not sure if this is possible but I'm looking to run say N
> simulations and record the p-val each time in order to estimate a power
> for the tests to pick up something that is there ... Is there any
> quicker way of doing this, rather than copying and pasting each p-value
> seperately in excel?
>
>
> Sorry for so many questions! Any help would be much appreciated.
> Tom.
>
>
>
>
>
>
>
>
>
>
>
> Quoting "Scott Morton, Fiona" <[email protected]>:
>
>> Yes, I use it all the time. Send any questions you have by email or
>> try me in the morning at +33 5 61 12 86 31
>> Fiona
>>
>> -----Original Message-----
>> From: TA Quilter [mailto:[email protected]]
>> Sent: Wed 2/1/2006 8:37 AM
>> To: Scott Morton, Fiona
>> Cc:
>> Subject: Stata?
>> Hi Fiona,
>>
>> I'm on of Ed's PhD students in my 3rd year.
>>
>> I was wondering if you'd ever used the stats package Stata in your
>> research and could spare me some mins for some quick questions?
>>
>> It would be a real help.
>>
>> Cheers,
>>
>> Tom.
>>>>
>>>>
>>>>
>>>> Quoting Richard Holt <[email protected]>:
>>>>
>>>>> Tom, Haibo and Marco,
>>>>>
>>>>> I attach exercises and solutions for Econ 2 this week and next week.
>>>>> Apologies for the timing - I have only just completed the solutions
>>>>> sheet.
>>>>>
>>>>> This week the exercise is based around analysis of economic data. I
>>>>> attach the exercise, the solution set and an Excel spreadsheet
>>>>> containing the data and the regression and other output.
>>>>>
>>>>> Students were asked to use Excel to look at histograms, compute
>>>>> descriptive statistics and run regressions (using consumption and
>>>>> income data (cross-section)).
>>>>>
>>>>> Most students have used Excel before - those intending to have the
>>>>> option to take a course containing economics (or accounting or business
>>>>> studies) at honours level have to take Computing in Management and
>>>>> Economics which includes a unit on Excel.
>>>>>
>>>>> As I understand it they did some probability theory with Jozsef last
>>>>> semester but this will be the first time that they look at statistics
>>>>> and / or do any data work. You might emphasise links between
>>>>> probability theory (esp mathematical expectation, correlation etc) and
>>>>> the sample counterparts.
>>>>>
>>>>> I have not devoted any time to statistical theory / methods in the
>>>>> lectures (we are currently discussing consumption / savings decisions
>>>>> under uncertainty, and I mention the idea of regression in order to be
>>>>> able to present Hall's work on the rational expectations permanent
>>>>> income model.
>>>>>
>>>>> The reason for setting an Excel based exercise is that they will be
>>>>> doing a project later this semester leading to a poster presentation,
>>>>> that will 'require' use of Excel to analyse and convey information
>>>>> about the relationship(s) between economic variables.
>>>>>
>>>>> I have asked students to bring their Excel output / a report containing
>>>>> details of the output to the tutorial for discussion.
>>>>>
>>>>> If you run out of things to discuss / discussion is not very fruitful
>>>>> then you could talk about the use of statistical analysis in economics,
>>>>> the difference between the population and sample distribution. The
>>>>> distinction between descriptive and inferential statistics, to
>>>>> illustrate that there is more to econometrics than computing means and
>>>>> standard deviations.
>>>>>
>>>>> If all else fails you can discuss material that you didn't have time to
>>>>> cover last week, or discuss issues that arise from this week's lectures.
>>>>>
>>>>> Let me know if you think the exercises/solutions contain errors so that
>>>>> I can correct them. Also could you let me know how the material is
>>>>> being received by the students - is it too difficult for them?
>>>>>
>>>>> Kind regards,
>>>>>
>>>>> Ric
>>>>>
>>>>>
>>>>
>>>>
>>>>
>>>>
>>>
>>>
>>>
>>>
>>
>>
>>
>>
>>
>>
>>
>>
>
>
>
>
>
>
>
>
*
* 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/