Talk:Development of Standard Scripts for Analysis and Programming

From PHUSE Wiki
Jump to: navigation, search

Why is there a section on a draft white paper for TFLs embedded in this "home page" for WG5? -- Mikec 15:35, 17 September 2012 (CDT)

This page contains text about:

Notice: Draft white paper on Tables, Flags and Listings associated with central tendency open for comments

which can't be edited out of this! I think this is a mistake and should be eliminated. Not sure who has admin privs, though.

Re: Why is there a section on a draft white paper for TFLs embedded in this "home page" for WG5? -- Sgetzin 14:04, 31 October 2012 (CDT)

Replace this text with your reply

Link has been fixed, please try again. Thanks, Scott and Mary

Sally Cassells - need to investigate why nobody seems to be able to edit the Discussion page.

From Harry Southworth at AZ (via email) I think there is an important question around whether central tendency is of any real interest for safety data. In general, it is the outliers that are of interest and looking at central tendency can actually be misleading - see the troglitazone example in Section 1.1 of Southworth & Heffernan (2012). Even if the centre is of interest, the mean is a terrible estimator for skewed data with outliers (and don't let me start on standard deviations). Even if it is Gaussian (and I've seen cases in which it is not even with N > 150), it is massively inefficient and robust alternatives exist. harrys

Comments to TFL associated with measures of central tendancy - focus on VS, ECG and Lab document -- Asca 02:20, 14 December 2012 (CST)

In general my company really appreciate the document and look forward to more of these. Especially for smaller companies it will be useful with standard recommendations on TFLs. A few comments:

6.1.2: Baseline should always be included when comparing post-baseline measurements.

6.1.3: The message in this section is not fully clear.

6.2.2: Reporting on the scale of measurement is reasonable provided that the correct summary statistics are included (e.g geometric mean and CV for distributions that are positively skewed as is often the case with lab parameters). Tests should take into account the distribution of the variable in question - this is not clearly stated.

7.1: The tables on change from minimum baseline to minumum post-baseline may be difficult to interpret. Similar commment on the maximum baseline to maximum post-baseline value. Perhaps some more text around interpretation would be helpful.

Åsa Carlsheimer on behalf of Ferring statisticians and programmers