0

I was kindly wondering whether you could shed a light onto the following concerns.

  1. I have a data where there are 10-12 independent variable (Heavy metals) against 1 dependent variable (being hazardous or non-hazardous).
  2. The sample size would be up to 25-30 I would like to ask whether I could use logit model to analyse the significance of heavy metals on being hazardous. I am afraid that the sample size is quite small but what would you suggest?

Thank you for all!

Jonpromie1
  • 35
  • 1
    I am not fully sure I understand your question. You have 10-12 independent variables that represent some metals (the presence or absence? or the amount present? or what type of variables are they?) and each observation corresponds to a material that contains such metals? Please try to explain more understandably how your data is (numeric/categorical/dichotomous) and what each observation represents. – Anna SdTC Aug 30 '17 at 09:36
  • 1
    Thank you for your fast answer. These 10-12 independent variables represent the presence of a heavy metal. For instance Pb, Fe, Zn etc. And values of these variables are numeric and represent the amount of each variable in this sample. To give a solid example: Pb - 0.01 mg/l ----- Zn- 0.03 mg/l ----- like this 10-12 independent variables... In the end we have one dependent variable ---- output- hazardous or non hazardous. Binary variable – Jonpromie1 Aug 30 '17 at 11:30

1 Answers1

2

Logit is the right model, but the sample size is probably too small. Perhaps try first with less independent variables in the hope you find at least something (this is called stepwise regression). I guess you know that in particular for small sample sizes you need to use the deviance statics with a chi-square test to test whether a variable is significant rather than just trusting the "asterisks" of your software output generated with a Wald test.

Tom Pape
  • 640