Re: Need Advice on a Simple Stat Question

On Mar 27, 6:05 am, "wki1...@xxxxxxxxxxx" <wki1...@xxxxxxxxxxx> wrote:
My statistics are a bit rusty and this is probably a problem for
statistics 101. I have daily data with each day having 2500 data
points, and I want to test if the average of the last 2 days is
significantly different from average of the previous 30 days.

The within day variances can be assumed equal.

What is the best test to use for this? The data is in SQL Server and
I'm using VB to raise a warning if there is a significant difference
in the last 2 days.

You could apply something like a student's t-test for dependent
samples, but I fear that with such large sample sizes you're going to
yield statistically significant results even for very small
differences. Perhaps a control chart would be more suitable in this
situation?

http://en.wikipedia.org/wiki/T-test

http://en.wikipedia.org/wiki/Control_chart
.

Relevant Pages

  • Re: Need Advice on a Simple Stat Question
    ... statistics 101. ... I have daily data with each day ... If the total variance is dominated by the variance between days, then you can base your statistical analysis on the daily averages, whether it's a two-sample t test or a linear regression. ...
    (sci.stat.math)
  • Need Advice on a Simple Stat Question
    ... My statistics are a bit rusty and this is probably a problem for ... I have daily data with each day having 2500 data ... I'm using VB to raise a warning if there is a significant difference ...
    (sci.stat.math)
  • Re: Effect Size and Sample Size
    ... You seem to be asking the technical, computing question. ... The smallest-N limit is passed if you divide by zero. ... There is a useful distinction that can be drawn between test statistics that are used, namely, "exact" and "approximate". ... With sample sizes of 2, there is not nearly as much room for departure from those assumptions as there is with sample sizes of 20 or more. ...
    (sci.stat.edu)
  • Re: Dataset Arrays or...?
    ... statistics over various time periods (e.g. monthly stats). ... Daily data will be in the form: ...
    (comp.soft-sys.matlab)
  • Re: Judaism, evolution, and science in general
    ... >ey.markov@xxxxxxxxx (Yisroel Markov) writes: ... >the latter is the fact of the statistics ... >>>scores within these racial categories are much greater than the variation ... >significant difference between two groups, ...
    (soc.culture.jewish.moderated)