As the title suggests, in general we are dealing with incomplete data.
For example, among personal information entered by customers, how many cases are the company name, position, department name, address, etc. correctly entered? One%? If you have data of 1 million people, you can say that the data of 10,000 people have been entered incorrectly.
Let's break down the example a bit more and talk about the position in the example above. The easiest way to access it (which is easy but will require quite a lot of manual work) is:
One. Outputs the distribution of positions from 1 million data
2. Check one by one >> For example, "employee" is " cool Isn't it wrongly marked as "," broken "Isn't marked with "...
3. After checking everything, write a simple (?) program. If you are using a SAS program, it would look like this:
if Position = "Siwon" then Position_1="Employee";
else if position="breakdown" then position_1="manager";
else if .....; /*I have to repeat it several times, right?*/
else Position_1=Title;
After executing the above program, check the data (distribution) of position / position_1, and if there is no problem, position_1 is used.
4. But... If another data is created/added, the above process is repeated. You should.
What about an address? " Gangnam-gu "Not" Kidney ", " Dongdaemun-gu "Not" Dongdamungu "... And many other unimaginable typos.
These cases can also appear in the field of medicine. When registering eCRF, fortunately, unless you enter the disease name in a drop-down method (in fact, it would be a tremendous effort to click and drop-down a number of diseases one by one, and find and click it would be tremendous effort). There is not.
70-80% of data analysis can be referred to as data preparation (including pre-processing steps). I dare say that how quickly and accurately you can perform these preparatory steps will determine the quality of your analysis work. Among the JMP Recode functions Fixing Misspelled Words You can achieve the purpose mentioned through.
Seoul address DB Example of correcting typos using
First of all, it is assumed that the following data exists in the DB. As you can see, the first 4 data are erroneously entered data, and the remaining 4 data are without typos.
If there is no JMP Recode function... In the example above, in order to change the first 4 typos, we need programming that corrects each typo. But, what if you use JMP? To make this easier Seoul address DB Receive and create a JMP data table in advance (see screenshot below, also provided as an attachment).
Everything is ready. Try the following (a brief video is also attached at the end).
One. Select the city, county and district variables from the JMP Data Table entered in error (mouse click)
2. JMP >> Cols >> Recode
3. Click RTM (Red Triangle Menu) in the Recode window >> Advanced >> Choose Closest
4. Choose Table >> Select Seoul
5. OK after selecting a city or district
6. Modifications completed
[Video] Correction of typos using JMP Recode function
