Can some one please help me in understanding how autocorrelation results in estimated standard errors tend to underestimate the true standard error.

# Query on blog - regression analysis and assumptions

Hi @divyark,

Please share the link to the blog you are mentioning here. Right now it’s difficult to understand the query.

Hi @divyark,

This is my understanding of the statement -

So if the error terms have some relation, then there is a high chance that the model missed some important factor while predicting. So the standard error will not be the true error.

Aishwarya ,

Could you help me further : i understand pattern in error terms which may be due to some pattern missed by the model to be captured in data and hence true error is different from standard error .

" If the error terms are correlated, the **estimated standard errors tend to underestimate the true standard error."**

When true standard error is lesses than the estimated standard error : What does this statement mean?

Aishwarya,

could you help me on the Query plz.

Regards,

Divya

I understand that you are clear with the previous explanation. Your next question is not clear to me. I believe this is the question-

I could not find the statement in the article, I could not locate the subtopic you are referring to. please mention which assumption are you talking about or under which heading can i find this line so that i can try to explain

Aishwarya : Let me re-iterate the Q: In the context below as copied from the blog

**2. Autocorrelation:** The presence of correlation in error terms drastically reduces model’s accuracy. This usually occurs in time series models where the next instant is dependent on previous instant. If the error terms are correlated, the estimated standard errors tend to underestimate the true standard error.

I understand in above referred text : presence of correlation in error terms means that there is a pattern which is not captured by the model and hence model accuracy is low. I did not understand the point " how the estimated standard errors tend to underestimate the true standard error" my understanding is that the estimated standard error will be higher than true standard errors ".

That’s right actually. Since we have missed out on some important relation and our errors show a pattern, chances are that the standard error calculated in this case would be more than the true standard error.