johnfergusonNUIG
commented
4 months ago Hi,
Thanks for providing this package.
when I use PAF_calc_discrete, it works perfectly with my dataset. However, with PAF_calc_continuous it doesnt't work and I get the following error: Error in impact_fraction(model = model, data = data, new_data = new_data, : Prevalence weighted estimation not appropriate for survival data sets In addition: Warning message: In risk_quantiles(riskfactor = riskfactor_vec[i], data = data, model = model, : NAs introduced by coercion
I cant provide the dataset since it's not allowed by the third party, but here is part of my code:
`df_bmi <- fread("paf_mi_bmi5.csv")
model <- with(df_bmi , coxph(Surv(TTE, event) ~ age +sex+ height+ active_cat2+ active_cat3+alc_cat2+ alc_cat3+ alc_cat4+ alc_cat5+ alc_cat6+ veg_imp+ fruit_imp+ crc_screen_imp+ depriv_imp+ smostat_cat2+ smostat_cat3+ packyears+ redmeat_cat2+ redmeat_cat3+ redmeat_cat4+ educ_cat2+ educ_cat3+ ethnic_bin2+ fh_crc_imp+ nsaids_imp+ processedmeat_cat2+ processedmeat_cat3+ processedmeat_cat4+ BMI5))
PAF_calc_continuous( model=model, riskfactor_vec=c("BMI5"), q_vec = c(0.01), data=df_bmi, calculation_method = "D", prev= 0.013, ci = TRUE, boot_rep = 200, t_vector = c(14.8), ci_level = 0.95, ci_type = c("norm"), S = 1, weight_vec = NULL, verbose = TRUE )
prev here is the prevalence of the disease (event) which is colorectal cancer.`
I'd appreciate any suggestions.
Hi Fatemeh,
As the help file says, the prevalece setting is appropriate for case-control datasets where you'd like an estimate of the disease prevalence avoided if a risk factor were eliminated (if you are using the setting method="D"). For survival data, the PAF esitmates the proportion of events that would be avoided at a particular time, in the absense of the risk factor. Since the data itself can estimate the proportion of events that might happen over various time intervals, you don't need to specify the prev argument. Try running without the prev argument...
John
FatemehSafi
Hi,
Thanks for providing this package.
when I use PAF_calc_discrete, it works perfectly with my dataset. However, with PAF_calc_continuous it doesnt't work and I get the following error: Error in impact_fraction(model = model, data = data, new_data = new_data, : Prevalence weighted estimation not appropriate for survival data sets In addition: Warning message: In risk_quantiles(riskfactor = riskfactor_vec[i], data = data, model = model, : NAs introduced by coercion
I cant provide the dataset since it's not allowed by the third party, but here is part of my code:
`df_bmi <- fread("paf_mi_bmi5.csv")
model <- with(df_bmi , coxph(Surv(TTE, event) ~ age +sex+ height+ active_cat2+ active_cat3+alc_cat2+ alc_cat3+ alc_cat4+ alc_cat5+ alc_cat6+ veg_imp+ fruit_imp+ crc_screen_imp+ depriv_imp+ smostat_cat2+ smostat_cat3+ packyears+ redmeat_cat2+ redmeat_cat3+ redmeat_cat4+ educ_cat2+ educ_cat3+ ethnic_bin2+ fh_crc_imp+ nsaids_imp+ processedmeat_cat2+ processedmeat_cat3+ processedmeat_cat4+ BMI5))
PAF_calc_continuous( model=model, riskfactor_vec=c("BMI5"), q_vec = c(0.01), data=df_bmi, calculation_method = "D", prev= 0.013, ci = TRUE, boot_rep = 200, t_vector = c(14.8), ci_level = 0.95, ci_type = c("norm"), S = 1, weight_vec = NULL, verbose = TRUE )
prev here is the prevalence of the disease (event) which is colorectal cancer.`
I'd appreciate any suggestions.