I am having trouble understanding what method to apply for the analysis of the following type of data:
id period grade GLP NAB ROE ROA OSS ESS PMA P30
1 1 NA 1790835 769 0.3795033 0.1171393 1.66014 1.221868 0.3976413 0.007312334
1 2 NA 2597351 1143 0.5896583 0.188034 1.947474 1.748581 0.4865142 0.01906484
1 3 15 3359203 1343 0.474556 0.1559151 1.814949 1.712795 0.4490202 0.02821528
2 3 16 12271110 15960 0.05 0.01 1.04 1.04 0.04 0.02
2 1 NA 5392109 9290 -0.02 -0.01 0.96 0.7 -0.04 0.02
2 2 NA 11128069 1253 -0.03332067 -0.005665337 0.9908606 0.9871047 -0.00922374 0.01818852
3 2 NA 820903 2225 0.06835087 0.004860066 1.012239 0.9919102 0.01209093 0.072
...
9 2 NA 652028 364 0.027 0.007 1.027 0.982 0.026 0.008
10 2 NA 31731236 251352 0.07033455 0.01930156 1.11506 1.069495 0.1031871 0.01566419
10 3 20 33016098 247747 0.0750367 0.02032998 1.124112 1.086046 0.1104093 0.01443777
10 1 NA 25151189 217818 0.03529102 0.01133722 1.059813 1.052836 0.0564377 0.02853068
They refer to credit ratings computed over three periods. The dependent variable is the grade (therefore ordered) and it does also depend on periods 1 and 2, but it not assigned until period 3, and therefore is missing. I thought to use ordered GEEs with a logit link, but in that case I would have either to put in the grade vector 0 or the period 3 values for each subject to analyze everything. Do you have any idea how to maintain the past observation without 'falsifying' the data? Help would be much appreciated!
