[Katharina Rogen]

Dear Mr. Schechter,

thank you so much! That was very helpful! I estimated it again -with the table option- and from what I make of it, the biggest difference is in the lowest group (it would expect about 1 case to have a yes and I have 9 cases), apart from that, the differences are relatively small (0-6 cases).

As the table option does not run with svy: i wrote a little program to feed the results from the svy logit model in a regular logit model and used the option then. If that´s ok and in case anyone else wants to try that this is the program I used

capture prog drop logit_post

    prog logit_post, eclass
    version 13
  
    mat b1= e(b)
    mat V1= e(V)
    
    loc com= e(command)
    qui: `com'
    ereturn repost b=b1 V=V1, rename
    end 

and this is my table

Logistic model for t_s, goodness-of-fit test

(Table collapsed on quantiles of estimated probabilities)
+--------------------------------------------------------+
Group    Prob  Obs_1  Exp_1  Obs_0  Exp_0  Total
-------+--------+-------+-------+-------+-------+-------
1  0.0103      9    0.9    115  123.1    124
2  0.0144      4    1.5    119  121.5    123
3  0.0183      2    2.0    122  122.0    124
4  0.0222      5    2.5    118  120.5    123
5  0.0270      4    3.1    120  120.9    124
-------+--------+-------+-------+-------+-------+-------
6  0.0313      4    3.6    119  119.4    123
7  0.0360      7    4.2    117  119.8    124
8  0.0404      2    4.7    121  118.3    123
9  0.0452      4    5.3    119  117.7    123
10  0.0514      6    6.0    118  118.0    124
-------+--------+-------+-------+-------+-------+-------
11  0.0579      3    6.7    120  116.3    123
12  0.0660      6    7.7    118  116.3    124
13  0.0755      6    8.7    117  114.3    123
14  0.0864      8   10.0    116  114.0    124
15  0.0980      9   11.3    114  111.7    123
-------+--------+-------+-------+-------+-------+-------
16  0.1089      6   12.8    117  110.2    123
17  0.1240     12   14.5    112  109.5    124
18  0.1406      8   16.3    115  106.7    123
19  0.1609     14   18.8    110  105.2    124
20  0.1856     16   21.3    107  101.7    123
-------+--------+-------+-------+-------+-------+-------
21  0.2115     25   24.7     99   99.3    124
22  0.2463     32   28.0     91   95.0    123
23  0.2908     39   32.9     84   90.1    123
24  0.3390     40   39.3     84   84.7    124
25  0.4028     43   45.6     80   77.4    123
-------+--------+-------+-------+-------+-------+-------
26  0.4913     46   55.0     78   69.0    124
27  0.5989     79   67.1     44   55.9    123
28  0.7335     85   82.8     39   41.2    124
29  0.8711     94   98.5     29   24.5    123
30  0.9999    113  115.9     10    7.1    123
+--------------------------------------------------------+

number of observations =      3703
number of groups =        30
Hosmer-Lemeshow chi2(28) =       109.26
Prob > chi2 =         0.0000 

Thank you & all the best,

Katharina

[amir habibdoost]

Hello everybody,
I am working on dyadic data (network data). I want to compute gof. But, as I know and as I read in FAQ, we should not do a likelihood-ratio test after a logit estimation with clustering.
I should not that I employed NGREG which wrote by MARCEL FAFCHAMPS. NGREG corrects standard errors in dyadic regressions.
A link about LR test in this condition:
https://www.stata.com/support/faqs/s...od-ratio-test/

Could someone please answer me and help me how can I compute gof?
Thank you

[jas esm]

Hello everyone,

I have a similar question and was wondering if anyone could please help me with it. I am running Cox regression models for a survey data and I can't run LRT or AIC to find the best fit for my models after this code: svy linearized : stcox predictor

Thank you.