SDM calculation fix

[anthonysena]

          Following up on this based on a question I received. Per https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3472075/, for continuous variables it appears we should divide the SD on line 139:

FeatureExtraction/R/CompareCohorts.R

Lines 134 to 141 in 6a7e90a

	m <- merge(covariates1, covariates2, all = T)
	m$mean1[is.na(m$mean1)] <- 0
	m$sd1[is.na(m$sd1)] <- 0
	m$mean2[is.na(m$mean2)] <- 0
	m$sd2[is.na(m$sd2)] <- 0
	m$sd <- sqrt(m$sd1^2 + m$sd2^2)
	m$stdDiff <- (m$mean2 - m$mean1) / m$sd
	result <- bind_rows(result, m[, c("covariateId", "mean1", "sd1", "mean2", "sd2", "sd", "stdDiff")])

similar to what we are doing for binary covariates here on line 103:

FeatureExtraction/R/CompareCohorts.R

Lines 96 to 105 in 6a7e90a

	m <- merge(covariates1, covariates2, all = T)
	m$count1[is.na(m$count1)] <- 0
	m$count2[is.na(m$count2)] <- 0
	m$mean1 <- m$count1 / n1
	m$mean2 <- m$count2 / n2
	m$sd1 <- sqrt(m$mean1 * (1 - m$mean1))
	m$sd2 <- sqrt(m$mean2 * (1 - m$mean2))
	m$sd <- sqrt((m$sd1^2 + m$sd2^2) / 2)
	m$stdDiff <- (m$mean2 - m$mean1) / m$sd
	result <- bind_rows(result, m[, c("covariateId", "mean1", "sd1", "mean2", "sd2", "sd", "stdDiff")])

Tagging @schuemie for confirmation. I can then open an issue and get this patched.

Originally posted by @anthonysena in #120 (comment)

[ginberg]

@anthonysena thanks for reporting, I will pick this up