Confidence Interval for Weighted Average Price Increases : Valid Approach?

[marcon33]

My goal is to measure a price increase from month to month at the invoice level and to include a valid confidence interval for each month to month comparison. I am taking 30 samples of 20,000 records out of 600K records and calculating the weighted average price increase for each sample. Ex. Unit Demand held constant x Old Price then x New price and the weighted average increase.
sample 1 , 20,000 records, $235000 extended value at old price, $239,700 value at new price = 2% increase

The 30 sample increases pass a test for normality. Can I say my confidence interval is valid?

Thank you in advance for any ideas....total rookie at this

[marcon33]

The best answer I found from a friend who is a six sigma black belt and math major was the need to perform two sample t tests and power analysis in order to define my sample size and power.

[hlsmith]

You probably could add a little more information to your posts to help us understand what you want to do and did. Are you trying to show a significant increase or trend?

Also, were you trying to link these two statements: "The 30 sample increases pass a test for normality. Can I say my confidence interval is valid?"?

You don't really discuss your confidence intervals. Did you construct confidence intervals around the sample averages or the differences in sequential averages? Are you saying, should I perform a t-test between monthly averages to show they are significantly different?

[marcon33]

Thanks , the question i am asking is what was the price increase for each month of the year based on invoice transactions. I want to measure the probable price increase as a percent comparing each monthe of the year to a price level last year in the 4th quarter. I am visualizing plotting theses monthly data points including a confidence interval and overlaying this on some industry data and tracking monthly.
Yes to linking those two statements. I have 15 samples of invoices (matched pairs at part number) where the weighted average price increase has been calculated. The population has 600k transactions and I have been pulling 5k matched pairs from the two time periods, 15 times. I calculated the CI for these 15 weighted averages after finding them to be normally distributed. Is the resulting CI reasonably valid.

The t test was unclear my apologies. After looking at a paired power t test in R ( pwr.t.test) I want to use this to calculate the sample size I should have defined earlier. Is 15 the right number of weighted average samples and what number of transactions are necessary for each of those samples. (the 5000 I am randomly pulling is arbitrary)
Thanks