Specialized AI agents for statistical analysis

As easy as ChatGPT, but local

Open a folder and ask in plain language. Blade runs R on your computer, so tables, plots, and the script come back quickly.

Pricing

Study A CHD risk factors

Study A › CHD risk factors

framingham.csv4,240 rows · 16 columns

age	sysBP	BMI	TenYearCHD
39	106	26.97	0
46	121	28.73	0
48	127.5	25.34	0
61	150	28.58	1
46	130	23.1	0
43	180	30.3	0

Which risk factors relate most to ten-year CHD?

Thinking

Shell.ExecRscript framingham_stats.RDone

Age and systolic BP are the strongest, then hypertension and diastolic BP.

with TenYearCHD	r	95% CI	p
age	0.23	0.20 to 0.25	< .001
sysBP	0.22	0.19 to 0.24	< .001
prevalentHyp	0.18	0.15 to 0.21	< .001
diaBP	0.15	0.12 to 0.17	< .001

Supervised

Study A CHD risk factors

Study A › CHD risk factors

R 4.4RunningRan in 0.4 s

1# Framingham -- linear regression output (OLS)2df <- read.csv("framingham.csv", check.names = FALSE)3 4# 2. multiple: sysBP5m2 <- lm(sysBP ~ age + male + BMI + heartRate + totChol,6         data = df)7print(summary(m2))

Console

Call:lm(formula = sysBP ~ age + male + BMI + heartRate + totChol,     data = df) Residuals:    Min      1Q  Median      3Q     Max -64.300 -12.481  -2.465   9.921 125.865  Coefficients:             Estimate Std. Error t value Pr(>|t|)    (Intercept) 22.531293   3.091221   7.289 3.72e-13 ***age          0.864168   0.035665  24.230  < 2e-16 ***male        -1.128182   0.596856  -1.890   0.0588 .  BMI          1.425642   0.072987  19.533  < 2e-16 ***heartRate    0.290839   0.024571  11.837  < 2e-16 ***totChol      0.036349   0.006852   5.305 1.19e-07 ***---Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1 Residual standard error: 18.84 on 4165 degrees of freedom  (69 observations deleted due to missingness)Multiple R-squared:  0.2642,	Adjusted R-squared:  0.2633 F-statistic: 299.1 on 5 and 4165 DF,  p-value: < 2.2e-16

	male	age	education	currentSmoker	cigsPerDay	BPMeds	prevalentHyp	totChol	sysBP	diaBP	BMI	heartRate	glucose	TenYearCHD
1	1	39	4	0	0	0	0	195	106	70	26.97	80	77	0
2	0	46	2	0	0	0	0	250	121	81	28.73	95	76	0
3	1	48	1	1	20	0	0	245	127.5	80	25.34	75	70	0
4	0	61	3	1	30	0	1	225	150	95	28.58	65	103	1
5	0	46	3	1	23	0	0	285	130	84	23.1	85	85	0
6	0	43	2	0	0	0	1	228	180	110	30.3	77	99	0
7	0	63	1	0	0	0	0	205	138	71	33.11	60	85	1
8	0	45	2	1	20	0	0	313	100	71	21.68	79	78	0
9	1	52	1	0	0	0	1	260	141.5	89	26.36	76	79	0
10	1	43	1	1	30	0	1	225	162	107	23.61	93	88	0
11	0	50	1	0	0	0	0	254	133	76	22.91	75	76	0
12	0	43	2	0	0	0	0	247	131	88	27.64	72	61	0
13	1	46	1	1	15	0	1	294	142	94	26.31	98	64	0
14	0	41	3	0	0	1	1	332	124	88	31.31	65	84	0
15	0	39	2	1	9	0	0	226	114	64	22.35	85	NA	0
16	0	38	2	1	20	0	1	221	140	90	21.35	95	70	1
17	1	48	3	1	10	0	1	232	138	90	22.37	64	72	0
18	0	46	2	1	20	0	0	291	112	78	23.38	80	89	1
19	0	38	2	1	5	0	0	195	122	84.5	23.24	75	78	0
20	1	41	2	0	0	0	0	195	139	88	26.88	85	65	0

==============================================================================
0. SAMPLE
==============================================================================
Rows: 4240   Columns: 16

==============================================================================
1. SIMPLE LINEAR REGRESSION: systolic BP ~ age
==============================================================================

Call:
lm(formula = sysBP ~ age, data = df)

Residuals:
    Min      1Q  Median      3Q     Max 
-57.382 -13.627  -2.627  10.182 148.042 

Coefficients:
            Estimate Std. Error t value Pr(>|t|)    
(Intercept) 82.14199    1.82570   44.99   <2e-16 ***
age          1.01276    0.03628   27.91   <2e-16 ***
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Residual standard error: 20.25 on 4238 degrees of freedom
Multiple R-squared:  0.1553,	Adjusted R-squared:  0.1551 
F-statistic:   779 on 1 and 4238 DF,  p-value: < 2.2e-16


Equation: sysBP = 82.142 + 1.0128 * age
n = 4240, R2 = 0.1553, adj R2 = 0.1551, residual SE = 20.253 mmHg
95% CI for slope: [0.9416, 1.0839]

==============================================================================
2. MULTIPLE LINEAR REGRESSION: sysBP ~ age + male + BMI + heartRate + totChol
==============================================================================

Call:
lm(formula = sysBP ~ age + male + BMI + heartRate + totChol, 
    data = df)

Residuals:
    Min      1Q  Median      3Q     Max 
-64.300 -12.481  -2.465   9.921 125.865 

Coefficients:
             Estimate Std. Error t value Pr(>|t|)    
(Intercept) 22.531293   3.091221   7.289 3.72e-13 ***
age          0.864168   0.035665  24.230  < 2e-16 ***
male        -1.128182   0.596856  -1.890   0.0588 .  
BMI          1.425642   0.072987  19.533  < 2e-16 ***
heartRate    0.290839   0.024571  11.837  < 2e-16 ***
totChol      0.036349   0.006852   5.305 1.19e-07 ***
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Residual standard error: 18.84 on 4165 degrees of freedom
  (69 observations deleted due to missingness)
Multiple R-squared:  0.2642,	Adjusted R-squared:  0.2633 
F-statistic: 299.1 on 5 and 4165 DF,  p-value: < 2.2e-16


n = 4171, R2 = 0.2642, adj R2 = 0.2633, F = 299.07, residual SE = 18.840

Ask about your data

Open CSV, Excel, or SPSS files and tell Blade what you want to know.

Read the results

Get tables, plots, and assumption checks in one place.

Keep the R script

Review the code, run it again, or share it with a colleague.

A workspace for each study

Keep data, conversations, and R scripts together. Switch projects without losing your place.

Blade Desktop

Study A

Study A 4 files

framingham.csvCSV · 192 KB2m
framingham_analysis.RR · 6 KB2m
R_fig3_correlation.pngPNG · 258 KB2m
framingham_analysis.pngPNG · 360 KB1h

Research is hard. The statistics shouldn't be

Your data took months to collect. The analysis shouldn't.

Where 588 clinical studies got stuck:

44.3%
Patient recruitment
26.7%
Data collection
21.5%
Data analysis and interpretation
20.4%
Study design
20.2%
Publication

Each square is 1% of the 588 studies. Cullati et al., BMC Medical Research Methodology, 2016

Trusted by leaders

For business decisions and research that needs clear answers.

Skala

“Honestly, our data, analysis, and R scripts used to be all over the place. Now everything lives in Blade, so the whole team can just open it and check the results together.”