Help calculating 95% confidence interval?

You’re probably trying to get excel to divide by zero, which is a math no-no. try using an IF(logical_argument; if_yes; if_no) function to determine if a certain cell has a value of 0, then follow through.

I kind of figured the same, but my SD is 0.. so what do I do?

dunno. but you can’t divide by 0. so, you could say IF SD = 0 then do nothing otherwise, divide by [cell reference] or something to that effect? i’m not a pro with statistics or anything, though…

Try changing your standard deviation to .00001 and see if your calculations work. The change between that and 0 will be insignificant in your final outcomes.

@fireinthepriory good call.

Thanks so much.. I tried putting in .00001 and the number I get is 8.00152E-06. Eh?

For those sets in which your SD is 0, I’m assuming this is because all of the numbers in the set are the same (otherwise, you have encountered some error when calculating your SD). Performing statistics on a set of numbers that are all the same doesn’t really work. Because there’s no distribution of values, there’s nothing to work with. Whatever 95% confidence interval you could possibly derive by some convoluted calculation wouldn’t be representative of anything, and it would be better to say that it was simply not applicable to that set of data.

@girlofscience Headdesk OF COURSE. Haha. A 95% confidence interval of a set of the same numbers doesn’t exist! You’re 100% confident because you have no variation. Haha. That’s what I get for only reading the Excell problem and not the statistics. :)

Yeah, it’s a small data set and all the numbers are the same. I have to go on to make a chart… so what do I do now?

The instructions say: “Construct a bar graph that displays the mean of each treatment for each variable. Place 95% confidence (error) bars on the means. The purpose here is that if the confidence whiskers overlap for two treatments then the treatments do not differ statistically from one another with regard to the response variable.”

Yep. Cool.

A set of data that is all the same number would not have error bars.

Okay, thanks. :)

No problem; just graph the bars for that represent the means of each treatment, and for the ones that have a standard deviation (i.e., they have a distribution of numbers in the set), add the error bars. For the treatments that have the same numbers across the set, there are no error bars. In this hypothetical and unrealistic problem, there are no error bars because there is no variance, so there is no statistical question that the mean is the only mean that could be achieved under the conditions. So, as long as the error bars from another treatment do not overlap with the mean of this treatment, they’re technically significantly different, in this strange data set you have. There are likely power issues because you have a small data set (but that’s a different issue entirely), and you generally can’t analyze anything that doesn’t have variance. But in this circumstance, I would simply plot all the treatments and include error bars only on the ones with a distribution of numbers.