On the Chi-Square Test When the Parameters are Estimated Independently of the Sample
Gerald R. Chase · Journal of the American Statistical Association · 1972
If the parameters are estimated independently of the sample, the chi-square test statistic for a goodness of fit test has a limiting distribution that is stochastically larger than that of the test of fit for a completely specified distribution. Thus, if the critical values for the test of fit for a completely specified distribution are incorrectly used, the probability that we will reject the null hypothesis when it is true is greater than the desired level of significance.