> The model fit for those eligible for the shingles vaccine predicts the risk of dementia decreases with age. Come on.
Except that's not what the chart he included shows.
1. That chart shows a strongly positive age correlation in the non-vaccine-eligible group: dementia diagnoses go up as you get closer to the left edge of the chart. X axis is birthdate, not age, so the left side is older.
2. The right half of the chart shows a relatively flat relationship between dementia diagnoses and age, very slightly negative age correlation. It's plausible that more data would show a slightly positive age correlation with dementia. Visually, it appears that if you omit the 6th point on the right side of the cutoff date (assuming that it's an outlier), the trend would be weakly positive with age.
3. The charts show _new dementia diagnoses_, not total rate of dementia, and I don't think it's actually that unreasonable to think that your risk of a new diagnosis might go down with age beyond a certain point. Assuming that dementia is a permanent condition (i.e. once you get it, you have it until you die; probably close enough to true for this argument), anyone previously diagnosed with dementia would be ineligible for a future new diagnosis. So long as the diagnosis rate is positive, your probability of having dementia for a given age will be higher for higher ages. As an analogy, your risk of a new diagnosis of type 1 diabetes goes down after about age 14, yet it's not illogical to expect that a higher proportion of 20-year-olds have T1D than 14-year-olds.
I don't think the slope of the line is damning but if there are outliers that throw the slope that much off, that may be a sign there's not enough data.
Also if you change the slope the effect size diminishes because the left side of the line tips up.
> The model fit for those eligible for the shingles vaccine predicts the risk of dementia decreases with age. Come on.
Except that's not what the chart he included shows.
1. That chart shows a strongly positive age correlation in the non-vaccine-eligible group: dementia diagnoses go up as you get closer to the left edge of the chart. X axis is birthdate, not age, so the left side is older.
2. The right half of the chart shows a relatively flat relationship between dementia diagnoses and age, very slightly negative age correlation. It's plausible that more data would show a slightly positive age correlation with dementia. Visually, it appears that if you omit the 6th point on the right side of the cutoff date (assuming that it's an outlier), the trend would be weakly positive with age.
3. The charts show _new dementia diagnoses_, not total rate of dementia, and I don't think it's actually that unreasonable to think that your risk of a new diagnosis might go down with age beyond a certain point. Assuming that dementia is a permanent condition (i.e. once you get it, you have it until you die; probably close enough to true for this argument), anyone previously diagnosed with dementia would be ineligible for a future new diagnosis. So long as the diagnosis rate is positive, your probability of having dementia for a given age will be higher for higher ages. As an analogy, your risk of a new diagnosis of type 1 diabetes goes down after about age 14, yet it's not illogical to expect that a higher proportion of 20-year-olds have T1D than 14-year-olds.
I don't think the slope of the line is damning but if there are outliers that throw the slope that much off, that may be a sign there's not enough data.
Also if you change the slope the effect size diminishes because the left side of the line tips up.