How do I clean my data?

[Omotola Oladapo]

Hi there, I need help with my data. After downloading my data from Google Forms, I imported it to Stata but my variables labels are showing as the questions. What can I do, please!

[Jared Greathouse]

Hey Omotola. Glad to see you've finally joined Statalist. Please, give an example dataset of your data using the dataex command. Specifically, show us the output of

dataex, varlabel

so that we may see your dataset exactly as you see it. There's an easy solution to this, but, you have to help us... help you.

[Ken Chui]

1) The variable label can be written over with another label. For example, assuming the desired label is "Respondent's gender":

label variable Whatgenderdoyouidentifyw "Respondent's gender"

2) The very long variable name can be difficult to type later during the analysis. It can also be renamed to something shorter like "gender":

rename Whatgenderdoyouidentifyw gender

3) What's more concerning is that it's a string variable with as many as 54 spaces. It means that some of the responses are much longer than "Female", as you can see there are totally 13 different "Unique values" in this variable. This means you'd have to run a tabulation using:

tabulate Whatgenderdoyouidentifyw

and may have to revise this variable into something that is easier to analyze.

To conclude, this is going to take a lot of time to investigate and clean. In future, questions with limited options like this would be better set up as a multiple choice in online questionnaires than an open-ended write-in question.

[Omotola Oladapo]

Hey Omotola. Glad to see you've finally joined Statalist. Please, give an example dataset of your data using the dataex command. Specifically, show us the output of

dataex, varlabel

so that we may see your dataset exactly as you see it. There's an easy solution to this, but, you have to help us... help you.

Thank you so much, for sharing this platform with me. I am excited I can get answers to my question!

[Omotola Oladapo]

1) The variable label can be written over with another label. For example, assuming the desired label is "Respondent's gender":

label variable Whatgenderdoyouidentifyw "Respondent's gender"

2) The very long variable name can be difficult to type later during the analysis. It can also be renamed to something shorter like "gender":

rename Whatgenderdoyouidentifyw gender

3) What's more concerning is that it's a string variable with as many as 54 spaces. It means that some of the responses are much longer than "Female", as you can see there are totally 13 different "Unique values" in this variable. This means you'd have to run a tabulation using:

tabulate Whatgenderdoyouidentifyw

and may have to revise this variable into something that is easier to analyze.

To conclude, this is going to take a lot of time to investigate and clean. In future, questions with limited options like this would be better set up as a multiple choice in online questionnaires than an open-ended write-in question.

Thank you Ken, I will try this steps now.

[Omotola Oladapo]

Hey Omotola. Glad to see you've finally joined Statalist. Please, give an example dataset of your data using the dataex command. Specifically, show us the output of

dataex, varlabel

so that we may see your dataset exactly as you see it. There's an easy solution to this, but, you have to help us... help you.

[Jared Greathouse]

Do what I wrote above, but include only 4 variables (dataex). Do not use screenshots, we can barely read them, and they don't help us help you

[Omotola Oladapo]

1) The variable label can be written over with another label. For example, assuming the desired label is "Respondent's gender":

label variable Whatgenderdoyouidentifyw "Respondent's gender"

2) The very long variable name can be difficult to type later during the analysis. It can also be renamed to something shorter like "gender":

rename Whatgenderdoyouidentifyw gender

3) What's more concerning is that it's a string variable with as many as 54 spaces. It means that some of the responses are much longer than "Female", as you can see there are totally 13 different "Unique values" in this variable. This means you'd have to run a tabulation using:

tabulate Whatgenderdoyouidentifyw

and may have to revise this variable into something that is easier to analyze.

To conclude, this is going to take a lot of time to investigate and clean. In future, questions with limited options like this would be better set up as a multiple choice in online questionnaires than an open-ended write-in question.

This solved it, I really appreciate Ken. The instruction was that there must be an open-ended question and a contingent question in the survey.

[Omotola Oladapo]

Do what I wrote above, but include only 4 variables (dataex). Do not use screenshots, we can barely read them, and they don't help us help you

I used the code again, and the error message was that the r(1000) input statement exceeded the line size limit. Try specifying fewer variables.

[Omotola Oladapo]

I am using the regress Craving gender but I got the below error. I apologize for attaching the screenshot.What do I need to do

[Andrew Musau]

Show us the result of

describe

and

dataex Cravings gender

[Ken Chui]

I am using the regress Craving gender but I got the below error. I apologize for attaching the screenshot.What do I need to do [ATTACH=CONFIG]n1709772[/ATTACH]

It is probably because the gender variable is still in string format (aka, it's entered as characters rather than being numerically coded). As I said before, string variables are often not accepted in many Stata commands. Regression (regress) is one of them. It will not accept string dependent or string independent variable. It will simply say "no observation."

My post in #3 is perhaps not explicit enough. In order to "clean" this data for it to make it ready for analysis, most of the variables will need to be turned into numerical form.

[Omotola Oladapo]

It is probably because the gender variable is still in string format (aka, it's entered as characters rather than being numerically coded). As I said before, string variables are often not accepted in many Stata commands. Regression (regress) is one of them. It will not accept string dependent or string independent variable. It will simply say "no observation."

My post in #3 is perhaps not explicit enough. In order to "clean" this data for it to make it ready for analysis, most of the variables will need to be turned into numerical form.

Thank you, Ken, I am new to data cleaning. Can you please explain how I can change the characters to numbers?

[Omotola Oladapo]

Show us the result of

describe

and

dataex Cravings gender

Andrew, this is the result of datax

[Omotola Oladapo]

I found online that I can use the encode command, [encode gender, generate(gender2)], and did this for all the variables. I was able to do the regress command. Thank you all for your support!