Some invalid responses in data

[Yaseminn Akcann]

Hi,

I am using a panel data where individuals are asked to answer their savingsinaccount and cash saving in terms of dollar amount. then I created total saving as a sum of these two. However, when I look at my data, I see that some values do not make sense in terms of savings. For instance if you look at input 745 savinginaccount is 1300268000 which is crazy. Or input 3092 where savinginaccount is 140000225000. What would be the best to do under this circumstances? I added both screenshots and data results. I am not willing to drop the observations so any help would be appreciated. Thank you!

input id float year double savinginaccount long cashsaving float totalsaving
129 1 450000 125000 575000
514 2 500000 400 500400
514 3 500000 500 500500
526 3 605000 5000 610000
711 3 500000 6000 506000
745 1 1300268000 200 1300268160
762 3 1450025 400 1450425
948 1 500000 10000 510000
948 2 700000 10000 710000
1226 1 2000000 50000 2050000
1507 1 288467 288467 576934
1543 2 300000 300000 600000
1561 3 818250 460 818710
1695 2 75000 300030000 300104992
1790 3 20000400 2500 20002900
2030 1 80001000 1200 80002200
2210 2 500000 20000 520000
2671 2 400000 400000 800000
2671 3 400000286 400000 400400288
3071 3 1000217 1300 1001517
3092 1 14000225000 500 14000225280
3289 1 2500000 5000 2505000
3441 3 15300950 500 15301450
3490 1 67000 1000000 1067000
3552 3 1500120 125 1500245
3570 1 500000 5000 505000
3685 1 500000 2500 502500
3744 1 700000 4000 704000
3744 2 900000 500 900500
3750 2 9300090 1100 9301190
3842 3 678000 39500 717500
3938 3 450000 100000 550000
4143 1 500000 5000 505000
4143 2 500000 1500 501500
4312 2 500000 35 500035
4327 3 10411041 400 10411441
4437 1 2000000 10000 2010000
4437 2 500000 10000 510000
4437 3 800000 10000 810000
4571 2 300060000 5000 300064992
4817 1 500 500000 500500
4838 3 500000 500000 1000000
5061 1 350000 350000 700000
5063 1 500000 200000 700000
5063 3 1000000 10000 1010000
5138 3 549000 500 549500
5635 1 12000000 500000 1.25e+07
5712 3 1000100 100 1000200
6013 3 603000 6000 609000
6158 2 120000 1000000 1120000
6348 3 4972106 1500 4973606
6553 3 2000000 2000000 4000000
6723 3 600000 1000 601000
6807 3 850000 300 850300
6819 3 530000 1000 531000
7498 2 1000000 500000 1500000
7523 3 1300000 0 1300000
7835 3 520000 75000 595000
7927 3 500000 500 500500
7958 3 -1000 600000 599000

[Carlo Lazzaro]

Yaseminn:
it would seem that the data entry was not that accurate.
For instance:

4327 3 10411041 400 10411441

sounds strange indeed.

I would send a query out to those who looked after that database setup asking for clarifications.
If your shot is unsuccessful, it's probably better to consider this values as missing and deal with them in the most appropriate way (-ipolate- or -mi-).

[William Lisowski]

I find the following two observations interesting.

2671 2 400000 400000 800000
2671 3 400000286 400000 400400288

Same ID in two years. The savingsaccount in year 3 is the same as the value in year two with "238" stuck onto the end. Clear some sort of data entry difficulty, or some problem importing the data into Stata. You might want to look at these two observations in your input data and see if there's anything to be learned.

Let me add that when you create totalsaving you will want to use

generate double totalsaving = savinginaccount + cashsaving

because the default if you don't specify the storage type is the same as if you used

generate float totalsaving = savinginaccount + cashsaving

and float cannot store integers larger than 16 million with complete accuracy, which is why in the year 3 observation above, totalsaving should have been 400400286 but was shown as 400400288.

Here are the limits on storage of decimal integers with full accuracy in the various numeric storage types. The fixed-point variables lose the 27 largest positive values to missing value codes; the similar loss for floating point variables occurs only for the largest exponent, so it doesn't affect the much smaller integer values.

byte - 7 bits	-127	100
int - 15 bits	-32,767	32,740
long - 31 bits	-2,147,483,647	2,147,483,620
float - 24 bits	-16,777,216	16,777,216
double - 53 bits	-9,007,199,254,740,992	9,007,199,254,740,992

[Yaseminn Akcann]

Thank you, Carlo and William! I ended up using these values as missing since my request for clarification was not answered.