non parametric ANCOVA

[Chiara Mar]

hi all,
I am working on my dissertation and I wanted to implement a non parametric ANCOVA (as in Wang and Akritas 2006 or Akritas, Arnold and Du 2000)
I want to investigate if group composition in term of gender has an effect on performance controlling for some covariates (both categorical and numerical). I have 3 groups (male, female, mixed).
I cannot apply ANCOVA because my dependent variable is not normally distributed. Do you have any idea on how to do that on stata?

help would be very much appreciated

thanks

chiara

[ericmelse]

Dear Chiara,

Possibly, this paper offers good advice to compare and examine median values between group categories:

The Stata Journal 2012 12 2 162-190 RM Conroy What hypotheses do nonparametric two-group tests actually test?

The method that Mr Conroy describes allows for the calculation of the median of your performance measure while controlling for covariates.

But, you could provide some example data that will make it possible to reply with syntax.

Best,
Eric

[Marcos Almeida]

Maybe you wish to take a look at - npregress - command.

[Chiara Mar]

Dear Chiara,

Possibly, this paper offers good advice to compare and examine median values between group categories:

The Stata Journal 2012 12 2 162-190 RM Conroy What hypotheses do nonparametric two-group tests actually test?

The method that Mr Conroy describes allows for the calculation of the median of your performance measure while controlling for covariates.

But, you could provide some example data that will make it possible to reply with syntax.

Best,
Eric

I have read the paper but I have not find out how to include covariates.

[Nick Cox]

#1

I am working on my dissertation and I wanted to implement a non parametric ANCOVA (as in Wang and Akritas 2006 or Akritas, Arnold and Du 2000)

Please note our request to avoid bare name and date references and give fuller details.

I want to investigate if group composition in term of gender has an effect on performance controlling for some covariates (both categorical and numerical). I have 3 groups (male, female, mixed).
I cannot apply ANCOVA because my dependent variable is not normally distributed. Do you have any idea on how to do that on stata?

The last statement does not bite as hard as you imply. Regression (it's an allowable quirk to call it ANCOVA) doesn't require a marginal normal distribution for the outcome, even as an ideal condition. At most a generalised linear model should suffice.

Why not post e.g. some details on the distribution of your performance measure? e.g. a dotplot and results of summarize, detail?

[Chiara Mar]

#1



Please note our request to avoid bare name and date references and give fuller details.



The last statement does not bite as hard as you imply. Regression (it's an allowable quirk to call it ANCOVA) doesn't require a marginal normal distribution for the outcome, even as an ideal condition. At most a generalised linear model should suffice.

Why not post e.g. some details on the distribution of your performance measure? e.g. a dotplot and results of summarize, detail?

dear Nick,
sorry for my mistake in including name.

I post some details on the distribution of performance (2 indicators are: time and last step)
I also post a part of the dataset:
1. gender_t1 is 2 if male, 1 if female, 0 is neither of the 2
2. knowledge is 1 if past experience, 0 is not
3. the 3 covariates I want to control for are Experience, age and GPA

[Nick Cox]

Please note our request to avoid bare name and date references and give fuller details.

That is explained in https://www.statalist.org/forums/help#references

Otherwise, thanks for showing more details.

Time seems to have an upper limit of 10. That needs to be explained and discussed.

Last step: You have one value of 0, between 5 and 10% of 1s and everything else 2. I wouldn't use that as anything but a predictor myself.