Open DanielGlorieux opened 4 months ago
Hi,
as any parametric model, the results are valid only if all assumptions are met. The violation of an hypothesis can lead to biased results, meaning that the estimations you get are not reliable. On real data, you can't know how far the true value is from your estimation, but you know that your results may be wrong.
Viviane
Thank you for your reply dear Viviane.
Taking the example of simple linear regression where certain assumptions may not be verified but do not prevent the use of the model (with some reservations), for example the hypothesis of homoscedasticity which when not respected leads to a loss of efficiency of the model. Still regarding simple linear regression, I saw here that violating the assumption of normality of residuals would lead to incorrect inferences for models with small samples but would not have much impact on model inference with large sample sizes. I was wondering if it would be the same for the lcmm model ?
As part of my study, I just use the lcmm model for the classification of my individuals, should I take all the hypotheses into account?
Thank you.
I am a student in the third year of a bachelor's degree. As part of a study I use the lcmm model, but the qq-plot graph of the residuals of my model are not normal. This is confirmed by Shapiro's test as well. What could be the consequences of the violation of this assumption on my study ? Thank you.