CRF 13: Continuous measurements > 100 for pancreatic ultrasound are probably missing (-1##) thus miscoded

[jsolon-ncp]

Upon inspecting the continuous measurements of USS ultrasounds, you find outliers at both ends. The negative values can be dealt with in the analysis .do files. However, the values > 100 probably have to be dealt with in the next cleaning of the master data. This is probably a -1## response.

stripplot pan_head_trans pan_head_ap pan_body_trans pan_tail_trans

shows the negative values which we can fix in the analysis .do file. There are also values > 100 for transverse readings of body and tail which should probably have been a -1##.

list sampa_id cohort pan_body_trans pan_tail_trans pan_head_* if pan_body_trans > 100 & pan_body_trans !=.

gives you this (which can be corrected in analysis .do file in the interim, but maybe at source as well.

[jsolon-ncp]

To find all rows with > 100 for uss pancreatic measurements and not missing (ie miscoded as + but missing)

foreach var of varlist pan_head_trans pan_head_ap pan_body_trans pan_tail_trans {
    list sampa_id cohort pan_head_trans pan_head_ap pan_body_trans pan_tail_trans if `var'>100 & `var'!=.
    }

[jsolon-ncp]

To fix at source :

foreach var of varlist pan_head_trans pan_head_ap pan_body_trans pan_tail_trans {
    replace `var' = `var'*(-1) if `var'>100 & `var'!=.
    }