Page 28 - CleanScience_Fall22

􏰀􏰁􏰂􏰀􏰃􏰄􏰂 􏰆􏰂 􏰇􏰈􏰃 􏰉􏰊􏰋􏰇 􏰇􏰃􏰂 􏰌􏰃􏰊􏰄􏰋 􏰍􏰆􏰇􏰈 􏰇􏰈􏰃 􏰎􏰂􏰏􏰆􏰂􏰐􏰋 􏰊􏰑􏰁􏰒􏰇 􏰇􏰈􏰃 􏰓􏰃􏰂􏰃􏰇􏰄􏰊􏰇􏰆􏰁􏰂 􏰁􏰔 􏰇􏰈􏰃 􏰒􏰉􏰇􏰄􏰊􏰕􏰎􏰂􏰃 􏰔􏰄􏰊􏰀􏰇􏰆􏰁􏰂 􏰁􏰔 􏰇􏰈􏰃 􏰖􏰗􏰘􏰙􏰚 􏰓􏰊􏰄􏰕 ticles through the lung tissues directly into the blood􏰕 stream. These particles are then distributed throughout the body, potentially affecting internal organs including the brain. Findings now include cardiovascular disease and an increasingly suspected role in the development of dementia and 􏰛l􏰜heimer􏰝s. 3, 4, 5, 6
Preliminary Associations of Dust and Reported Symptoms of Residential Occupants
As detailed below, three methods of logistic regressions were run on data involving dust as an exploration of potential associations via the comparisons of patterns from statistical analysis. The purpose and structure of the questionnaire was to discover relationships, so the data is observational and anecdotal, and we cannot infer causality. Because the information was essentially 􏰞crowd sourced,􏰟 the reported data are subjective. However, subjective information can still be useful if managed transparently, and is a useful approach to learning about how occupants of homes perceive their houses, their behaviors, and the potential effects on health of exposure and interventions.
Correlations can emerge and hypotheses developed from pattern comparisons, resulting in questions for further investigation. The large si􏰜e of the data set allows the opportunity for visibility into real􏰕world examples beyond that of other databases. Following are examples of three different regression methods of analysis involving dust.
Symptom Type Logistic Regressions
􏰠 􏰡urface dust is a statistically signi􏰎cant contributor to 􏰊􏰉􏰉 􏰘􏰢 􏰋􏰌􏰣􏰓􏰇􏰁􏰣􏰋 􏰤􏰓 􏰥 􏰦􏰙􏰦􏰦􏰘􏰧.
􏰠 􏰨n terms of cumulative 􏰩dds 􏰪atio 􏰤􏰩􏰪􏰧 for symptoms, surface dust 􏰤average 􏰩􏰪 􏰫 􏰬.3􏰭􏰧 is the second most impactful condition behind 􏰑􏰃􏰉􏰆􏰃􏰔 􏰇􏰈􏰊􏰇 􏰈􏰁􏰒􏰋􏰃 􏰆􏰋 􏰆􏰣􏰓􏰊􏰀􏰇􏰆􏰂􏰐 􏰈􏰃􏰊􏰉􏰇􏰈 􏰤􏰊􏰮􏰃􏰄􏰊􏰐􏰃 􏰩􏰪 􏰫 􏰬􏰙􏰯􏰯􏰧􏰙
􏰠 􏰡urface dust is a particularly strong contributor to 􏰔􏰊􏰇􏰆􏰐􏰒􏰃 􏰤􏰩􏰪 􏰫 􏰬.6􏰘􏰧, 􏰋􏰆􏰂􏰒􏰋 􏰤􏰩􏰪 􏰫 􏰬.5􏰭􏰧, 􏰊􏰀􏰈􏰆􏰂􏰐 􏰤􏰩􏰪 􏰫 􏰬.54􏰧, 􏰣􏰁􏰁􏰏 􏰀􏰈􏰊􏰂􏰐􏰃􏰋 􏰤􏰩􏰪 􏰫 􏰬.5􏰘􏰧, and 􏰀􏰁􏰒􏰐􏰈􏰆􏰂􏰐 􏰤􏰩􏰪 􏰫 􏰬.5􏰦􏰧.
Number of Symptoms Linear Regression
The same housing and behavioral characteristics utili􏰜ed in the logistic regressions were used in an 􏰩rdi􏰕 nary 􏰰east 􏰡quares 􏰤􏰩􏰰􏰡􏰧 linear regression attempting to predict total number of symptoms a respondent would report. The 􏰩􏰰􏰡 linear regression reported an R2 of 0.243 (Table 2).
􏰠 􏰡urface 􏰱ust was the number 􏰬 predicter of number of symptoms (p < 0.000001) (Table 3).
􏰠 Respondents who indicated 􏰡urface 􏰱ust report 0.506 more symptoms than those who do not report Surface Dust (Table 3).
􏰠 􏰲otice 􏰩dors and Sill Dust were the number 2 and number 3 predictors of number of symptoms respectively (p < 0.000001) (Table 3).
Dep. Variable
Model
Method
No. Observations
Df Residuals
Df Model
R-squared
Adj. R-squared
F-statistic
Number of symptoms
OLS
Least Squares
55536
55360
175
0.243
0.241
101.6
Prob (F-statistic))
Table 2: Number of Symptoms Linear Regression Model
Summary
0.0%
Table 3: Top 10 feature effects for the Number of Symptoms Linear Regression
28 | The Cleaning Industry Research Institute FALL 2022