http://www.dphu.org/uploads/attachements/books/books_3134_0.pdf

Does this help ...

https://www.youtube.com/watch?v=siqx4PbqJ6s

or

https://support.microsoft.com/en-gb/help/828124/description-of-the-confidence-statistical-functions-in-excel

You have a few questions. It sounds like the primary one is how to calculate CIs in Excel for predicted values and it looks like that is covered in the previous responses.

I'll take a stab at the second, what happens to standard errors and CIs when outliers are deleted. Assuming outliers are ones that don't fit the pattern, removing these will decrease the standard error and thus the width of the CIs (and increase the likelihood of significance). If the outliers are removed because they are not ones that should be included in the model (the example I give is perception scientists dealing with reaction times less than a second and a participant sneezes for answers their mobile phone), it is less problematic (you can assume their phone ring happens at random), but if you are removing value just because they are outliers (so called trimming) you should use a different equation for you standard errors. Rand Wilcox has several textbooks (and papers) that cover this really well.

Sometimes deleting outliers discards information that you want. See the attached paper. Best, D. Booth PS Robust Statistics is much better than removing outliers.

Dear all,

Thank you for your answers, and your time.

I think that the answer will be:

Confidence interval for expected value = expected value +- t.value* standard error

Where t.value = TINV(propability,df) . df resulted from removing outlines. Standard error is included in the regression analysis table.

Best Regards,

Shaimaa