[R] rms package: output interpretation

[David Winsemius]

> On Mar 25, 2016, at 6:31 AM, T.Riedle <tr206 at kent.ac.uk> wrote:
> 
> Hi everybody,
> 
> I am trying to run a logistic regression using the rms package. Here is the output of my model.
> 
> Logistic Regression Model
> 
> lrm(formula = stock.market.crash ~ crash.t.1.to.t.L + MA.inflator.monthly +
>    realized.volatility.10 + MA.MP.100 + MA.UI.100 + MA.DEI.100 +
>    MA.UPR.100, data = FDL.model_monthly)
>                     Model Likelihood     Discrimination    Rank Discrim.
>                        Ratio Test            Indexes          Indexes
> Obs            45    LR chi2     21.57    R2       0.529    C       0.889
> 0             30    d.f.            7    g        2.789    Dxy     0.778
> 1             15    Pr(> chi2) 0.0030    gr      16.267    gamma   0.778
> max |deriv| 6e-06                         gp       0.340    tau-a   0.354
>                                          Brier    0.131
> 
>                       Coef      S.E.     Wald Z Pr(>|Z|)
> Intercept               -11.6543   5.9683 -1.95  0.0509
> crash.t.1.to.t.L         -4.5335   2.5705 -1.76  0.0778
> MA.inflator.monthly      -2.4400   1.2735 -1.92  0.0554
> realized.volatility.10   24.7952  10.3298  2.40  0.0164
> MA.MP.100                 6.9404   4.1511  1.67  0.0945
> MA.UI.100              -125.7101  54.5219 -2.31  0.0211
> MA.DEI.100              519.9589 255.0241  2.04  0.0415
> MA.UPR.100                2.6938   2.2209  1.21  0.2252
> 
> 
> I am a bit confused regarding the interpretation of the chi2 and its p-value. Can anybody help me interpret the results?

I don't think anyone can help you unless you first describe the structure of the data. I worry that you are analyzing some sort of panel structure and have not yet accounted for autocorrelation in monthly measures.


> I get a high R2 but chi2 seems to be significant and high. How do I interpret these results in the rms package?

This suggests you need to talk to someone with deeper statistical background, and that's not really what r-help bills itself as providing. Is this part of an education experience or task? Do you have a supervisor that could be consulted?


-- 

David Winsemius
Alameda, CA, USA