Dear all,

I have a panel dataset of 2700 firms with quarterly dates from 1985 to 2017. My panel is unbalanced. My dependent variable is ratings, and the independent variables are some financial ratios. The feologit command results in the desired way. But my problem is when I include the years' fixed effect, the likelihood won't converge. How should I look for a problem ?

If it helps, the distribution of rating data and the distribution of data in the sample are as follows:




rating Freq. Percent Cum.

1 124 0.15 0.15
2 147 0.18 0.34
3 428 0.53 0.87
4 857 1.07 1.94
5 2,601 3.25 5.19
6 5,143 6.42 11.61
7 9,139 11.41 23.01
8 9,782 12.21 35.22
9 7,730 9.65 44.87
10 5,448 6.80 51.67
11 7,303 9.12 60.79
12 9,444 11.79 72.58
13 6,125 7.64 80.22
14 4,622 5.77 85.99
15 5,619 7.01 93.00
16 2,638 3.29 96.30
17 1,139 1.42 97.72
18 1,130 1.41 99.13
19 259 0.32 99.45
20 440 0.55 100.00

Total 80,118 100.00



YEARLY DATE Freq. Percent Cum.

1985 420 0.52 0.52
1986 1,611 2.01 2.54
1987 1,782 2.22 4.76
1988 1,866 2.33 7.09
1989 1,790 2.23 9.32
1990 1,660 2.07 11.39
1991 1,642 2.05 13.44
1992 1,736 2.17 15.61
1993 1,902 2.37 17.98
1994 2,038 2.54 20.53
1995 2,168 2.71 23.23
1996 2,427 3.03 26.26
1997 2,557 3.19 29.46
1998 2,744 3.42 32.88
1999 2,771 3.46 36.34
2000 2,806 3.50 39.84
2001 3,032 3.78 43.63
2002 3,145 3.93 47.55
2003 3,122 3.90 51.45
2004 3,147 3.93 55.38
2005 3,063 3.82 59.20
2006 3,040 3.79 62.99
2007 2,996 3.74 66.73
2008 2,959 3.69 70.43
2009 2,957 3.69 74.12
2010 2,919 3.64 77.76
2011 2,899 3.62 81.38
2012 2,892 3.61 84.99
2013 2,843 3.55 88.54
2014 2,841 3.55 92.08
2015 2,890 3.61 95.69
2016 2,787 3.48 99.17
2017 666 0.83 100.00

Total 80,118 100.00