About P21C v4.1.0

Matt

on June 5, 2024

P21

Hi MinJi,   Thank you for…

Hi MinJi,

Thank you for posting. The images uploaded were removed by the system for security purposes. Can you please attach the files again? You can also send the images to community@pinnacle21.com for review and I can follow back up here.

Thanks,

Matt

Hi Matt,   Thank you for…

Hi Matt,

Thank you for follow up my posting.

Instead of uploading screenshots, I describe the issue: In orifinal SAS dataset, value is 152.1835366. But, in converted DatasetJSON, the value has been changed to 152.183537.

Thanks,

MinJi

Matt

on June 10, 2024

P21

Hi MinJi,   Thank you for…

Hi MinJi,

Thank you for the added details! On conversion, float values are expected to be truncated to 6 digits as that is the default precision for float values. In your example of 152.1835366, is this exact level of precision desired/necessary? I can provide the feedback you provide to our team who can evaluate changes to the precision.

Kind regards,

Matt

Dataset-JSON format number

A bit more and precise information ...

There can be two reasons for the (small) difference:
a) P21C "truncates" the values to 6 digits after the decimal point.
Though this can sound plausible, this may cause problems when the value is e.g. 0.0000366 ...
b) SAS-XPT uses another way of storing numeric values, i.e. the "IBM format", which was the most used format in the era of IBM mainframe computers (do you still have one?). This format is not used by modern computers anymore. This alone can already lead to a (small) difference in precision when converting from outdated "IBM format" to modern (IEEE) format, and there is not much one can do against it.

As part of the CDISC Dataset-JSON "hackathon" (https://cosa.cdisc.org/hackathons/datasetJson), several software tools were developed (mostly "open source") to generate Dataset-JSON, also starting from SAS-XPT. You might try out some of these to see whether one of these gives more (for you) satisfactory results.

Lex

on June 10, 2024

Truncation

It would interesting to know if in SAS 152.1835366 was an exact value, or formatted, or truncated?

What is the “real” number?

In Dataset-JSON 1.1 you will be able to represent numbers as decimal strings that can be converted than to numbers upon import.

Thank you for the response. …

Thank you for the response.

@Lex,

The value is from "Area under the concentration-time curve from time zero to infinity (extrapolated)" in external data and imported as is into SDTM without format.

@Jozef, 

The e-6 is not a small difference compared to the e-16 or e-17, so I want to reduce the difference. And, since the DatasetJSON conversion function was added to P21C, I wanted to use it.

It was converted to retain decimal points using one of the open source codes.Thanks

@Matt,

According to the FDA Rule ID SD1212 (FDAB031), Standardized Result in Numeric Format (--STRESN) variable value should be equal Standardized Result in Character Format (--STRESC) variable value, when Standardized Result in Character Format (--STRESC) variable value represents a numeric value.

That's why I want to keep the original decimal places.

Matt

on June 12, 2024

P21

Hello MinJi,   Thank you…

Hello MinJi,

Thank you for the added context regarding the concerns with SD1212. I have escalated the questions to our internal team who will review and follow up with me, and then I will relay messaging in our thread here!

Kind regards,

Matt