# T-test Help

Six Sigma – iSixSigma Forums Old Forums General T-test Help

Viewing 4 posts - 1 through 4 (of 4 total)
• Author
Posts
• #44400

Brit
Participant

I have a company who provides survey data.  They use the t-test (assuming unequal variances) and have the following explanation of their results:
When a score on a report is found to be significantly different from the score on the last report, the score is highlighted with one of the following symbols:
+          95% certainty of significant increase (t >1.96).
+ +       99% certainty of significant increase (t >2.576).
–           95% certainty of significant decrease (t< -1.96).
– –         99% certainty of significant decrease (t< -2.576).
Given that sample sizes and variances are different from report to report, does anyone have a problem with their reasoning?  Wondering why they do not use the test statistic and compare against the values from the t-test table given degrees of freedom.  I understand the z-table relationship, but think I would get different results.
Help!

0
#142129

BTDT
Participant

Brit:I have two points:1) They are using Z-scores and saying they are t-scores. The two are equivalent only for infinite sample size (is this true?).2) The explanation implies a single tailed test, but the critical values are those for a two-tailed test.3) There are just some people who are bad with numbers.Cheers, BTDT

0
#142134

Brit
Participant

They are sampling from a patient population, so I believe the pool is not infinite.
My conclusion was the same as your #3 below.

0
#142139

BTDT
Participant

Brit:I have seen some hospitals use control charts for reporting such things as number of C. difficile cases. They weren’t too bad at giving a visual picture of the statistics. They can satisfy the statistically adroit and would be better suited to combat the “one off” mentality of only comparing results to the previous month…. just a thought.Cheers, BTDT

0
Viewing 4 posts - 1 through 4 (of 4 total)

The forum ‘General’ is closed to new topics and replies.