[Andreas Rydell]

After exclusion of those individuals with afib-diagnos at baseline I want to know the following:

Time from examination date (examdate1) to diagnose (ICD10), where I want the earliest if those cases there are more than one. rowmin should be there somewhere I think. Last time rowmin only showed the lowest value, not the lowest value in those columns with I48 only.
afibdd=afib diagnose date

gen afibdd=0
egen earliest = rowmin(ts_41280_0_*)
format earliest %td
forvalues i=0/242 {
local var s_41270_0_`i'
local time ts_41280_0_`i'
recode afibdd (0=1) if `time'==earliest & substr(`var', 1, 3) =="I48"
}

Doesn't search for earliest time in those with I48 only like I want.

. tab afib afibdd

| afibdd
afib | 0 1 | Total
-----------+----------------------+----------
0 | 986 0 | 986
1 | 50 5 | 55
-----------+----------------------+----------
Total | 1,036 5 | 1,041

[Andreas Rydell]

I found the code here on statalist for time to event from baseline:

gen ttdm = (earliest - examdate1)/(365/12)

As mentioned in previous post I need an if statement when making earliest. In the output now I get minus values, wich wouldn't be possible since I have excluded all individuals that have diagnose prior to baseline examination.

[Nick Cox]

#15 would seem equivalent to

gen afib = 0 

foreach var of varlist s_41270_0_* {
    replace `var' = "I48" if substr(`var', 1, 3) =="I48" 
    replace afib = 1 if `var' == "I48"
}

generate exclude = 0 

foreach time of var ts_41280_0_* {  
    replace exclude = 1 if afib == 1 & `time' <= examdate1
}

tab exclude
tab afib

[Andreas Rydell]

#15 would seem equivalent to

gen afib = 0

foreach var of varlist s_41270_0_* {
replace `var' = "I48" if substr(`var', 1, 3) =="I48"
replace afib = 1 if `var' == "I48"
}

generate exclude = 0

foreach time of var ts_41280_0_* {
replace exclude = 1 if afib == 1 & `time' <= examdate1
}

tab exclude
tab afib

The previous code had 65 with afib vs 10 exclude.

The code from #18 has 49 exclude. When I manually look in the data it seems that 10 is correct.

What I'm struggling with now is to get the earliest date if more than one column with I48 diagnose:

gen afibdd=0
egen earliest = rowmin (ts_41280_0_*)
format earliest %td
forvalues i=0/242 {
local var s_41270_0_`i'
local time ts_41280_0_`i'
recode afibdd (0=1) if (`time'==earliest & substr(`var', 1, 3) =="I48")
}
tab afib afibdd

gen ttdm = (earliest - examdate1)/(365/12)
gen ttdy = ttd/12
drop if earliest<=examdate1
tab earliest

If only earliest could be the date with I48 or earliest date if several, Im all set for calculating time to diagnos from baseline.

[Nick Cox]

We aren't seeing any real(istic) data examples but the earliest date with a I48 as a diagnosis may need to be sought with a customised loop.

gen earliest = . 

forvalues i=0/242 {
local var s_41270_0_`i'
local time ts_41280_0_`i'
replace earliest = `time' if `time' < earliest & substr(`var', 1, 3) =="I48")
}

[Andreas Rydell]

We aren't seeing any real(istic) data examples but the earliest date with a I48 as a diagnosis may need to be sought with a customised loop.

gen earliest = .

forvalues i=0/242 {
local var s_41270_0_`i'
local time ts_41280_0_`i'
replace earliest = `time' if `time' < earliest & substr(`var', 1, 3) =="I48")
}

Thank you. That worked. I now have time from baseline examination to afib diagnose among those with the diagnose only. For the other individuals there is =. in the column. I did two time to diagnose (months/year, TTDM and TTDY respectively). But I also need to make a time to censor for the other one to be able to do survival analyses:

replace ttdm = td(01jan2022-examdate1)/(365/12) if ttdy==. doesnt work in its current form. Jan 2022 was chosen just to have a follow up date.

I previously did:

gen ttdm = (earliest - examdate1)/(365/12)
*gen ttdy = ttd/12
drop if earliest<=examdate1

None were dropped like it should since those had been removed in a previous step.

[Nick Cox]

td(1jan2022) - examdate1

is more likely to be what you want.

[Andreas Rydell]

td(1jan2022) - examdate1

is more likely to be what you want.

Wonderful Nick. Thank you very much. Almost there now. Maybe you easily can see why my output gets wrong here?

I have time to event in month (ttdm) and years (ttdy) among those individuals that get an event from examination date to end of follow up. So far so good:

gen earliest = .
forvalues i=0/242 {
local var s_41270_0_`i'
local time ts_41280_0_`i'
replace earliest = `time' if (`time' < earliest & substr(`var', 1, 3) =="I48")
}
tab earliest
format earliest %td
gen ttdm = (earliest - examdate1)/(365/12)
gen ttdy = ttd/12

The output here is correct in month and year.


The problem is all those individuals that never get afib and has a "." in ttdm. I want an end date of follow up for survival analyses. For some reason I get wrong data output-format. I tried to tweak the code slightly with no success.

replace ttdm = td(01jan2022)-examdate1 if ttdm==.

ttdm
4764
4686
94.68493 correct
5343
4214
4679
145.0192 correct


replace ttdm = td(01jan2022)-examdate1/(365/12) if ttdm==.

same here,
ttdm
22595.62
89.42466
22596.1

I think it has to do with "td(01jan2022)" somehow since "(earliest-examdate1)/(365/12) get correct output.

[Nick Cox]

td(1jan2022) is just 22646. If you're getting correct answers in some cases and not others, it's either the data or the code. You seem to be subtracting something that's not a daily date from a daily date, which seems puzzling if not wrong.

[William Lisowski]

Your problem is that the expression to the right of the = in your replace command should be exactly like the expression in your generate command but with earliest changed to td(1jan2022)

[Andreas Rydell]

When I did like this it worked

gen enddate= td(01jan2022)
format enddate %td
replace ttdm = (enddate-examdate1)/(365/12) if ttdm==.

Can't thank you guys enough for your help.🙏

[Andreas Rydell]

Here we go again....

If these two works:

gen afib=0
foreach var of varlist s_41270_0_* {
recode afib (0=1) if `var'=="I48"
}

foreach var of varlist s_41270_0_* {
replace `var'="I48" if (`var'=="I480" | `var'=="I481" ///
| `var'=="I482" | `var'=="I483" | `var'=="I484" | `var'=="I485" ///
| `var'=="I486" | `var'=="I487" | `var'=="I488" | `var'=="I489")
}

Why do I get "typ mismatch" error on this?:

gen statintreatment=0
foreach var of varlist n_20003_0_* {
recode statintreatment (0=1) if `var'=="1141146234"

}

and "==invalid name" error on this?

foreach var of varlist n_20003_0_* {
replace `var'=="1141146234" if (`var'=="1141146138" | `var'=="1141192410" ///
| `var'=="1141192414" | `var'=="1140888648" | `var'=="1140861970" | `var'=="1140861958" ///
| `var'=="1140881748")
}

[William Lisowski]

Why do I get "type mismatch" error on this?:

gen statintreatment=0
foreach var of varlist n_20003_0_* {
recode statintreatment (0=1) if `var'=="1141146234"

My guess is that your variables beginning n_20003_0_ are numeric variables rather than string variables, so you should in theory have

recode statintreatment (0=1) if `var'==1141146234

I will add that for this to be correct it requires that your n_20003_0_ variables be either double (preferably) or long. You have 10-digit numbers; double will hold precisely numbers up to 15 digits long, while long will hold precisely numbers up to 9 digits long, or 10 digits no larger than 2,147,483,620.

and "==invalid name" error on this?

foreach var of varlist n_20003_0_* {
replace `var'=="1141146234" if (`var'=="1141146138" | `var'=="1141192410" ///
| `var'=="1141192414" | `var'=="1140888648" | `var'=="1140861970" | `var'=="1140861958" ///
| `var'=="1140881748")
}

Because "==" is used only for comparison, as in

`var'=="1141146138"

"=" is needed for generate and replace, as in

replace `var'="1141146234"

But once you get past that error, I expect the first problem will again occur, since you are again comparing and replacing the n_20003_0_ variables using strings rather than numbers.

[Andreas Rydell]

n2003 is type= long and 41270 is type=str6

I think this could be the reason it doesnt work.

[Andreas Rydell]

I wrote at the same time you were posting.

EDIT: I found a workaround.