determining statistical significance

**koochandkai** · 06-22-2012, 04:15 PM

I have datasets for a bunch of years and am comparing different metrics for the various years. I want to determine whether the change from one year to the next is statistically significant.
For example, in Year 1 there are 100 calls and in Year 2 there are 105 calls. I want to find out if the increase of 5 in the first two years is statistically significant.
Similarly, the average time in Year 1 is 2 minutes and the average time in Year 2 is 2.2 minutes. Is this .2 increase statistically significant.
I also calculate 90th percentiles. If 90% of the calls for Year 1 are below 3 minutes and 90% of the calls in Year 2 are below 3.1 minutes, I want to know if the .1 minute increase is significant.

I hope someone out there can help me!

Thanks so much.

**roki4** · 06-22-2012, 04:22 PM

What is your null and alternate hypothesis and p-value?

The data is statisically significant if this increase you used as an example did not occur by chance. That is, if the cause of the increase in call time is due to the fact that there are more steps in procedure or slower technology, etc, etc.

**koochandkai** · 06-22-2012, 04:32 PM

The statistics I learned long ago is starting to bubble up.
My null hypothesis would be that the average call times for the two years is the same. Yes? And the alternative is that they are not the same.
And then if I reject the null, I assume the alternative with 95% probability (if p is .05) -- yes?

**roki4** · 06-22-2012, 04:33 PM

Yep, you can do anywhere between a p of 0.10 and 0.01 if you choose to change your condfidence interval.

Edit: Are you also after things like...the average call time is 3 minutes with a standard deviation of 30 seconds. Meaning 99% of the time, calls fall between a duration of 1min30secs and 4mins30sec?

Just remember that 99.7% of data fall within 3 standard deviations of the mean.

**koochandkai** · 06-22-2012, 04:41 PM

I'm almost there -- or at least closer.....!

Is there an excel formula that will magically do this?

**roki4** · 06-22-2012, 04:47 PM

Doubt it, because you have to use a z-table to look up your different coefficients. I'm sure an add-on is available but if not you'll have to sit down with a pencil and paper.

Z-Table Here